Differentiable architecture search (DARTS) has signif-icantly promoted the development of NAS techniques because of its high search efficiency and effectiveness but suf-fers from performance collapse. In this paper, we make efforts to alleviate the performance collapse problem for DARTS from two aspects. First, we investigate the expres-sive power of the supernet in DARTS and then derive a new setup of DARTS paradigm with only training Batch-Norm. Second, we theoretically find that random features dilute the auxiliary connection role of skip-connection in supernet optimization and enable search algorithm focus on fairer operation selection, thereby solving the performance collapse problem. We instantiate DARTS and PC-DARTS with random features to build an improved version for each named RF-DARTS and RF-PCDARTS respectively. Experimental results show that RF-DARTS obtains 94.36% test accuracy on CIFAR-10 (which is the nearest optimal result in NAS-Bench-201), and achieves the newest state-of-the-art top-1 test error of 24.0% on ImageNet when transferring from CIFAR-10. Moreover, RF-DARTS performs robustly across three datasets (CIFAR-10, CIFAR-100, and SVHN) and four search spaces (S1-S4). Besides, RF-PCDARTS achieves even better results on ImageNet, that is, 23.9% top-1 and 7.1% top-5 test error, surpassing representative methods like single-path, training-free, and partial-channel paradigms directly searched on ImageNet.

Data augmentation is a critical technique in object detection, especially the augmentations targeting at scale invariance training (scale-aware augmentation). However, there has been little systematic investigation of how to design scale-aware data augmentation for object detection. We propose Scale-aware AutoAug to learn data augmentation policies for object detection. We define a new scale-aware search space, where both image- and instance-level augmentations are designed for maintaining scale robust feature learning. Upon this search space, we propose a new search metric, termed Pareto Scale Balance, to facilitate efficient augmentation policy search. In experiments, Scale-aware AutoAug yields significant and consistent improvement on various object detectors (e.g., RetinaNet, Faster R-CNN, Mask R-CNN, and FCOS), even compared with strong multi-scale training baselines. Our searched augmentation policies are generalized well to other datasets and instance-level tasks beyond object detection, e.g., instance segmentation. The search cost is much less than previous automated augmentation approaches for object detection, i.e., 8 GPUs across 2.5 days versus. 800 TPU-days. In addition, meaningful patterns can be summarized from our searched policies, which intuitively provide valuable knowledge for hand-crafted data augmentation design. Based on the searched scale-aware augmentation policies, we further introduce a dynamic training paradigm to adaptively determine specific augmentation policy usage during training. The dynamic paradigm consists of an heuristic manner for image-level augmentations and a differentiable copy-paste-based method for instance-level augmentations. The dynamic paradigm achieves further performance improvements to Scale-aware AutoAug without any additional burden on the long tailed LVIS benchmarks. We also demonstrate its ability to prevent over-fitting for large models, e.g., the Swin Transformer large model. Code and models are available at https://github.com/dvlab-research/SA-AutoAug.

In this paper, we present a conceptually simple, strong, and efficient framework for fully- and weakly-supervised panoptic segmentation, called Panoptic FCN. Our approach aims to represent and predict foreground things and background stuff in a unified fully convolutional pipeline, which can be optimized with point-based fully or weak supervision. In particular, Panoptic FCN encodes each object instance or stuff category with the proposed kernel generator and produces the prediction by convolving the high-resolution feature directly. With this approach, instance-aware and semantically consistent properties for things and stuff can be respectively satisfied in a simple generate-kernel-then-segment workflow. Without extra boxes for localization or instance separation, the proposed approach outperforms the previous box-based and -free models with high efficiency. Furthermore, we propose a new form of point-based annotation for weakly-supervised panoptic segmentation. It only needs several random points for both things and stuff, which dramatically reduces the annotation cost of human. The proposed Panoptic FCN is also proved to have much superior performance in this weakly-supervised setting, which achieves 82% of the fully-supervised performance with only 20 randomly annotated points per instance. Extensive experiments demonstrate the effectiveness and efficiency of Panoptic FCN on COCO, VOC 2012, Cityscapes, and Mapillary Vistas datasets. And it sets up a new leading benchmark for both fully- and weakly-supervised panoptic segmentation.

Homography estimation is a basic image alignment method in many applications. It is usually done by extracting and matching sparse feature points, which are error-prone in low-light and low-texture images. On the other hand, previous deep homography approaches use either synthetic images for supervised learning or aerial images for unsupervised learning, both ignoring the importance of handling depth disparities and moving objects in real-world applications. To overcome these problems, in this work, we propose an unsupervised deep homography method with a new architecture design. In the spirit of the RANSAC procedure in traditional methods, we specifically learn an outlier mask to only select reliable regions for homography estimation. We calculate loss with respect to our learned deep features instead of directly comparing image content as did previously. To achieve the unsupervised training, we also formulate a novel triplet loss customized for our network. We verify our method by conducting comprehensive comparisons on a new dataset that covers a wide range of scenes with varying degrees of difficulties for the task. Experimental results reveal that our method outperforms the state-of-the-art, including deep solutions and feature-based solutions.

In this paper, we propose a novel query design for the transformer-based object detection. In previous transformer-based detectors, the object queries are a set of learned embeddings. However, each learned embedding does not have an explicit physical meaning and we cannot explain where it will focus on. It is difficult to optimize as the prediction slot of each object query does not have a specific mode. In other words, each object query will not focus on a specific region. To solve these problems, in our query design, object queries are based on anchor points, which are widely used in CNN-based detectors. So each object query focuses on the objects near the anchor point. Moreover, our query design can predict multiple objects at one position to solve the difficulty: ``one region, multiple objects''. In addition, we design an attention variant, which can reduce the memory cost while achieving similar or better performance than the standard attention in DETR. Thanks to the query design and the attention variant, the proposed detector that we called Anchor DETR, can achieve better performance and run faster than the DETR with 10x fewer training epochs. For example, it achieves 44.2 AP with 19 FPS on the MSCOCO dataset when using the ResNet50-DC5 feature for training 50 epochs. Extensive experiments on the MSCOCO benchmark prove the effectiveness of the proposed methods. Code is available at https://github.com/megvii-research/AnchorDETR.

In this paper, we propose the first self-distillation framework for general object detection, termed LGD (Label-Guided self-Distillation). Previous studies rely on a strong pretrained teacher to provide instructive knowledge that could be unavailable in real-world scenarios. Instead, we generate an instructive knowledge by inter-and-intra relation modeling among objects, requiring only student representations and regular labels. Concretely, our framework involves sparse label-appearance encoding, inter-object relation adaptation and intra-object knowledge mapping to obtain the instructive knowledge. They jointly form an implicit teacher at training phase, dynamically dependent on labels and evolving student representations. Modules in LGD are trained end-to-end with student detector and are discarded in inference. Experimentally, LGD obtains decent results on various detectors, datasets, and extensive tasks like instance segmentation. For example in MS-COCO dataset, LGD improves RetinaNet with ResNet-50 under 2x single-scale training from 36.2% to 39.0% mAP (+ 2.8%). It boosts much stronger detectors like FCOS with ResNeXt-101 DCN v2 under 2x multi-scale training from 46.1% to 47.9% (+ 1.8%). Compared with a classical teacher-based method FGFI, LGD not only performs better without requiring pretrained teacher but also reduces 51% training cost beyond inherent student learning.

Non-uniformed 3D sparse data, e.g., point clouds or voxels in different spatial positions, make contribution to the task of 3D object detection in different ways. Existing basic components in sparse convolutional networks (Sparse CNNs) process all sparse data, regardless of regular or submanifold sparse convolution. In this paper, we introduce two new modules to enhance the capability of Sparse CNNs, both are based on making feature sparsity learnable with position-wise importance prediction. They are focal sparse convolution (Focals Conv) and its multi-modal variant of focal sparse convolution with fusion, or Focals Conv-F for short. The new modules can readily substitute their plain counterparts in existing Sparse CNNs and be jointly trained in an end-to-end fashion. For the first time, we show that spatially learnable sparsity in sparse convolution is essential for sophisticated 3D object detection. Extensive experiments on the KITTI, nuScenes and Waymo benchmarks validate the effectiveness of our approach. Without bells and whistles, our results outperform all existing single-model entries on the nuScenes test benchmark. Code and models are at github.com/dvlab-research/FocalsConv.

6D object pose estimation networks are limited in their capability to scale to large numbers of object instances due to the close-set assumption and their reliance on high-fidelity object CAD models. In this work, we study a new open set problem; the few-shot 6D object poses estimation: estimating the 6D pose of an unknown object by a few support views without extra training. To tackle the problem, we point out the importance of fully exploring the appearance and geometric relationship between the given support views and query scene patches and propose a dense prototypes matching framework by extracting and matching dense RGBD prototypes with transformers. Moreover, we show that the priors from diverse appearances and shapes are crucial to the generalization capability under the problem setting and thus propose a large-scale RGBD photorealistic dataset (ShapeNet6D) for network pre-training. A simple and effective online texture blending approach is also introduced to eliminate the domain gap from the synthesis dataset, which enriches appearance diversity at a low cost. Finally, we discuss possible solutions to this problem and establish benchmarks on popular datasets to facilitate future research. [project page]

Long-tailed instance segmentation is a challenging task due to the extreme imbalance of training samples among classes. It causes severe biases of the head classes (with majority samples) against the tailed ones. This renders “how to appropriately define and alleviate the bias” one of the most important issues. Prior works mainly use label distribution or mean score information to indicate a coarse-grained bias. In this paper, we explore to excavate the confusion matrix, which carries the fine-grained misclassification details, to relieve the pairwise biases, generalizing the coarse one. To this end, we propose a novel Pairwise Class Balance (PCB) method, built upon a confusion matrix which is updated during training to accumulate the ongoing prediction preferences. PCB generates fightback soft labels for regularization during training. Besides, an iterative learning paradigm is developed to support a progressive and smooth regularization in such debiasing. PCB can be plugged and played to any existing method as a complement. Experimental results on LVIS demonstrate that our method achieves state-of-the-art performance without bells and whistles. Superior results across various architectures show the generalization ability. The code and trained models are available at https://github.com/megvii-research/PCB.

In this work, we present a conceptually simple yet effective framework for cross-modality 3D object detection, named voxel field fusion. The proposed approach aims to maintain cross-modality consistency by representing and fusing augmented image features as a ray in the voxel field. To this end, the learnable sampler is first designed to sample vital features from the image plane that are projected to the voxel grid in a point-to-ray manner, which maintains the consistency in feature representation with spatial context. In addition, ray-wise fusion is conducted to fuse features with the supplemental context in the constructed voxel field. We further develop mixed augmentor to align feature-variant transformations, which bridges the modality gap in data augmentation. The proposed framework is demonstrated to achieve consistent gains in various bench-marks and outperforms previous fusion-based methods on KITTI and nuScenes datasets. Code is made available at https://github.com/dvlab-research/VFF11Part of the work was done in MEGVII Research..

Sparsely annotated semantic segmentation (SASS) aims to train a segmentation network with coarse-grained (i.e., point-, scribble-, and block-wise) supervisions, where only a small proportion of pixels are labeled in each image. In this paper, we propose a novel tree energy loss for SASS by providing semantic guidance for unlabeled pixels. The tree energy loss represents images as minimum spanning trees to model both low-level and high-level pair-wise affini-ties. By sequentially applying these affinities to the net-work prediction, soft pseudo labels for unlabeled pixels are generated in a coarse-to-fine manner, achieving dynamic online self-training. The tree energy loss is effective and easy to be incorporated into existing frameworks by com-bining it with a traditional segmentation loss. Compared with previous SASS methods, our method requires no multi-stage training strategies, alternating optimization proce-dures, additional supervised data, or time-consuming post-processing while outperforming them in all SASS settings. Code is available at https://github.com/megvii-research/TreeEnergyLoss.

Autonomous driving requires the model to perceive the environment and (re)act within a low latency for safety. While past works ignore the inevitable changes in the environment after processing, streaming perception is proposed to jointly evaluate the latency and accuracy into a single metric for video online perception. In this paper, instead of searching trade-offs between accuracy and speed like previous works, we point out that endowing real-time models with the ability to predict the future is the key to dealing with this problem. We build a simple and effective frame-work for streaming perception. It equips a novel Dual-Flow Perception module (DFP), which includes dynamic and static flows to capture the moving trend and basic detection feature for streaming prediction. Further, we introduce a Trend-Aware Loss (TAL) combined with a trend factor to generate adaptive weights for objects with different moving speeds. Our simple method achieves competitive performance on Argoverse-HD dataset and improves the AP by 4.9% compared to the strong baseline, validating its effectiveness. Our code will be made available at https://github.com/yancie-yjr/StreamYOLO.

In this paper, we propose a new query-based detection framework for crowd detection. Previous query-based detectors suffer from two drawbacks: first, multiple predictions will be inferred for a single object, typically in crowded scenes; second, the performance saturates as the depth of the decoding stage increases. Benefiting from the nature of the one-to-one label assignment rule, we propose a progressive predicting method to address the above issues. Specifically, we first select accepted queries prone to generate true positive predictions, then refine the rest noisy queries according to the previously accepted predictions. Experiments show that our method can significantly boost the performance of query-based detectors in crowded scenes. Equipped with our approach, Sparse RCNN achieves 92.0% AP, 41.4% MR−2and 83.2% JI on the challenging CrowdHuman [35] dataset, outperforming the box-based method MIP [8] that specifies in handling crowded scenarios. Moreover, the proposed method, robust to crowdedness, can still obtain consistent improvements on moderately and slightly crowded datasets like CityPersons [47] and COCO [26]. Code will be made publicly available at https://github.com/megvii-model/Iter-E2EDET.

In this work, we present a unified framework for multi-modality 3D object detection, named UVTR. The proposed method aims to unify multi-modality representations in the voxel space for accurate and robust single- or cross-modality 3D detection. To this end, the modality-specific space is first designed to represent different inputs in the voxel feature space. Different from previous work, our approach preserves the voxel space without height compression to alleviate semantic ambiguity and enable spatial connections. To make full use of the inputs from different sensors, the cross-modality interaction is then proposed, including knowledge transfer and modality fusion. In this way, geometry-aware expressions in point clouds and context-rich features in images are well utilized for better performance and robustness. The transformer decoder is applied to efficiently sample features from the unified space with learnable positions, which facilitates object-level interactions. In general, UVTR presents an early attempt to represent different modalities in a unified framework. It surpasses previous work in single- or multi-modality entries. The proposed method achieves leading performance in the nuScenes test set for both object detection and the following object tracking task. Code is made publicly available at https://github.com/dvlab-research/UVTR.

In this paper, we explore the mask representation in instance segmentation with Point-of-Interest (PoI) features. Differentiating multiple potential instances within a single PoI feature is challenging, because learning a high-dimensional mask feature for each instance using vanilla convolution demands a heavy computing burden. To address this challenge, we propose an instance-aware convolution. It decomposes this mask representation learning task into two tractable modules as instance-aware weights and instance-agnostic features. The former is to parametrize convolution for producing mask features corresponding to different instances, improving mask learning efficiency by avoiding employing several independent convolutions. Meanwhile, the latter serves as mask templates in a single point. Together, instance-aware mask features are computed by convolving the template with dynamic weights, used for the mask prediction. Along with instance-aware convolution, we propose PointINS, a simple and practical instance segmentation approach, building upon dense one-stage detectors. Through extensive experiments, we evaluated the effectiveness of our framework built upon RetinaNet and FCOS. PointINS in ResNet101 backbone achieves a 38.3 mask mean average precision (mAP) on COCO dataset, outperforming existing point-based methods by a large margin. It gives a comparable performance to the region-based Mask R-CNN K. He, G. Gkioxari, P. Dollár, and R. Girshick, "Mask R-CNN," in Proc. IEEE Int. Conf. Comput. Vis., 2017, pp. 2980-2988 with faster inference.

We propose a novel point annotated setting for the weakly semi-supervised object detection task, in which the dataset comprises small fully annotated images and large weakly annotated images by points. It achieves a balance between tremendous annotation burden and detection performance. Based on this setting, we analyze existing detectors and find that these detectors have difficulty in fully exploiting the power of the annotated points. To solve this, we introduce a new detector, Point DETR, which extends DETR by adding a point encoder. Extensive experiments conducted on MS-COCO dataset in various data settings show the effectiveness of our method. In particular, when using 20% fully labeled data from COCO, our detector achieves a promising performance, 33.3 AP, which outperforms a strong baseline (FCOS) by 2.0 AP, and we demonstrate the point annotations bring over 10 points in various AR metrics.

We propose a new convolution called Dynamic Region-Aware Convolution (DRConv), which can automatically assign multiple filters to corresponding spatial regions where features have similar representation. In this way, DR-Conv outperforms standard convolution in modeling semantic variations. Standard convolutional layer can increase the number of filers to extract more visual elements but results in high computational cost. More gracefully, our DR-Conv transfers the increasing channel-wise filters to spatial dimension with learnable instructor, which not only improve representation ability of convolution, but also maintains computational cost and the translation-invariance as standard convolution dose. DRConv is an effective and elegant method for handling complex and variable spatial information distribution. It can substitute standard convolution in any existing networks for its plug-and-play property, especially to power convolution layers in efficient networks. We evaluate DRConv on a wide range of models (MobileNet series, ShuffleNetV2, etc.) and tasks (Classification, Face Recognition, Detection and Segmentation). On ImageNet classification, DRConv-based ShuffleNetV2-0.5 achieves state-of-the-art performance of 67.1% at 46M× multiply-adds level with 6.3% relative improvement.

This paper revisits feature pyramids networks (FPN) for one-stage detectors and points out that the success of FPN is due to its divide-and-conquer solution to the optimization problem in object detection rather than multi-scale feature fusion. From the perspective of optimization, we introduce an alternative way to address the problem instead of adopting the complex feature pyramids - utilizing only one-level feature for detection. Based on the simple and efficient solution, we present You Only Look One-level Feature (YOLOF). In our method, two key components, Dilated Encoder and Uniform Matching, are proposed and bring considerable improvements. Extensive experiments on the COCO benchmark prove the effectiveness of the proposed model. Our YOLOF achieves comparable results with its feature pyramids counterpart RetinaNet while being 2.5× faster. Without transformer layers, YOLOF can match the performance of DETR in a single-level feature manner with 7× less training epochs. Code is available at https://github.com/megvii-model/YOLOF.

We present a simple but powerful architecture of convolutional neural network, which has a VGG-like inference-time body composed of nothing but a stack of 3 × 3 convolution and ReLU, while the training-time model has a multi-branch topology. Such decoupling of the training-time and inference-time architecture is realized by a structural re-parameterization technique so that the model is named RepVGG. On ImageNet, RepVGG reaches over 80% top-1 accuracy, which is the first time for a plain model, to the best of our knowledge. On NVIDIA 1080Ti GPU, RepVGG models run 83% faster than ResNet-50 or 101% faster than ResNet-101 with higher accuracy and show favorable accuracy-speed trade-off compared to the state-of-the-art models like EfficientNet and RegNet. The code and trained models are available at https://github.com/megvii-model/RepVGG.

Recently few-shot object detection is widely adopted to deal with data-limited situations. While most previous works merely focus on the performance on few-shot categories, we claim that detecting all classes is crucial as test samples may contain any instances in realistic applications, which requires the few-shot detector to learn new concepts without forgetting. Through analysis on transfer learning based methods, some neglected but beneficial properties are utilized to design a simple yet effective few-shot detector, Retentive R-CNN. It consists of Bias-Balanced RPN to debias the pretrained RPN and Re-detector to find few-shot class objects without forgetting previous knowledge. Extensive experiments on few-shot detection benchmarks show that Retentive R-CNN significantly outperforms state-of-the-art methods on overall performance among all settings as it can achieve competitive results on few-shot classes and does not degrade the base class performance at all. Our approach has demonstrated that the long desired never-forgetting learner is available in object detection.

Recent advances in label assignment in object detection mainly seek to independently define positive/negative training samples for each ground-truth (gt) object. In this paper, we innovatively revisit the label assignment from a global perspective and propose to formulate the assigning procedure as an Optimal Transport (OT) problem – a well-studied topic in Optimization Theory. Concretely, we define the unit transportation cost between each demander (anchor) and supplier (gt) pair as the weighted summation of their classification and regression losses. After formulation, finding the best assignment solution is converted to solve the optimal transport plan at minimal transportation costs, which can be solved via Sinkhorn-Knopp Iteration. On COCO, a single FCOS-ResNet-50 detector equipped with Optimal Transport Assignment (OTA) can reach 40.7% mAP under 1× scheduler, outperforming all other existing assigning methods. Extensive experiments conducted on COCO and CrowdHuman further validate the effectiveness of our proposed OTA, especially its superiority in crowd scenarios. The code is available at https://github.com/Megvii-BaseDetection/OTA.

In this work, we present FFB6D, a Full Flow Bidirectional fusion network designed for 6D pose estimation from a single RGBD image. Our key insight is that appearance information in the RGB image and geometry information from the depth image are two complementary data sources, and it still remains unknown how to fully leverage them. Towards this end, we propose FFB6D, which learns to combine appearance and geometry information for representation learning as well as output representation selection. Specifically, at the representation learning stage, we build bidirectional fusion modules in the full flow of the two networks, where fusion is applied to each encoding and decoding layer. In this way, the two networks can leverage local and global complementary in-formation from the other one to obtain better representations. Moreover, at the output representation stage, we designed a simple but effective 3D keypoints selection algorithm considering the texture and geometry information of objects, which simplifies keypoint localization for precise pose estimation. Experimental results show that our method outperforms the state-of-the-art by large margins on several benchmarks. Code and video are available at https://github.com/ethnhe/FFB6D.git.

In this paper, we present a conceptually simple, strong, and efficient framework for panoptic segmentation, called Panoptic FCN. Our approach aims to represent and predict foreground things and background stuff in a unified fully convolutional pipeline. In particular, Panoptic FCN encodes each object instance or stuff category into a specific kernel weight with the proposed kernel generator and produces the prediction by convolving the high-resolution feature directly. With this approach, instance-aware and semantically consistent prosperties for things and stuff can be respectively satisfied in a simple generate-kernel-then-segment workflow. Without extra boxes for localization or instance separation, the proposed approach outperforms previous box-based and -free models with high efficiency on COCO, Cityscapes, and Mapillary Vistas datasets with single scale input. Our code is made publicly available at https://github.com/Jia-Research-Lab/PanopticFCN.1

We present an unsupervised learning approach for optical flow estimation by improving the upsampling and learning of pyramid network. We design a self-guided upsample module to tackle the interpolation blur problem caused by bilinear upsampling between pyramid levels. Moreover, we propose a pyramid distillation loss to add supervision for intermediate levels via distilling the finest flow as pseudo labels. By integrating these two components together, our method achieves the best performance for unsupervised optical flow learning on multiple leading benchmarks, including MPI-SIntel, KITTI 2012 and KITTI 2015. In particular, we achieve EPE=1.4 on KITTI 2012 and F1=9.38% on KITTI 2015, which outperform the previous state-of-the-art methods by 22.2% and 15.7%, respectively.

We present a simple, effective, and general activation function we term ACON which learns to activate the neurons or not. Interestingly, we find Swish, the recent popular NAS-searched activation, can be interpreted as a smooth approximation to ReLU. Intuitively, in the same way, we approximate the more general Maxout family to our novel ACON family, which remarkably improves the performance and makes Swish a special case of ACON. Next, we present meta-ACON, which explicitly learns to optimize the parameter switching between non-linear (activate) and linear (inactivate) and provides a new design space. By simply changing the activation function, we show its effectiveness on both small models and highly optimized large models (e.g. it improves the ImageNet top-1 accuracy rate by 6.7% and 1.8% on MobileNet-0.25 and ResNet-152, respectively). Moreover, our novel ACON can be naturally transferred to object detection and semantic segmentation, showing that ACON is an effective alternative in a variety of tasks. Code is available at https://github.com/nmaac/acon.

We propose a dense object detector with an instance-wise sampling strategy, named IQDet. Instead of using human prior sampling strategies, we first extract the regional feature of each ground-truth to estimate the instance-wise quality distribution. According to a mixture model in spatial dimensions, the distribution is more noise-robust and adapted to the semantic pattern of each instance. Based on the distribution, we propose a quality sampling strategy, which automatically selects training samples in a probabilistic manner and trains with more high-quality samples. Extensive experiments on MS COCO show that our method steadily improves baseline by nearly 2.4 AP without bells and whistles. Moreover, our best model achieves 51.6 AP, outperforming all existing state-of-the-art one-stage detectors and it is completely cost-free in inference time.

Mainstream object detectors based on the fully convolutional network has achieved impressive performance. While most of them still need a hand-designed non-maximum suppression (NMS) post-processing, which impedes fully end-to-end training. In this paper, we give the analysis of discarding NMS, where the results reveal that a proper label assignment plays a crucial role. To this end, for fully convolutional detectors, we introduce a Prediction-aware One-To-One (POTO) label assignment for classification to enable end-to-end detection, which obtains comparable performance with NMS. Besides, a simple 3D Max Filtering (3DMF) is proposed to utilize the multi-scale features and improve the discriminability of convolutions in the local region. With these techniques, our end-to-end framework achieves competitive performance against many state-of-the-art detectors with NMS on COCO and CrowdHuman datasets. The code is available at https://github.com/Megvii-BaseDetection/DeFCN.

In this paper, we investigate a new variant of neural architecture search (NAS) paradigm – searching with random labels (RLNAS). The task sounds counter-intuitive for most existing NAS algorithms since random label provides few information on the performance of each candidate architecture. Instead, we propose a novel NAS framework based on ease-of-convergence hypothesis, which requires only random labels during searching. The algorithm involves two steps: first, we train a SuperNet using random labels; second, from the SuperNet we extract the subnetwork whose weights change most significantly during the training. Extensive experiments are evaluated on multiple datasets (e.g. NAS-Bench-201 and ImageNet) and multiple search spaces (e.g. DARTS-like and MobileNet-like). Very surprisingly, RLNAS achieves comparable or even better results compared with state-of-the-art NAS methods such as PC-DARTS, Single Path One-Shot, even though the counterparts utilize full ground truth labels for searching. We hope our finding could inspire new understandings on the essential of NAS.

Despite the recent success of deep neural networks, it remains challenging to effectively model the long-tail class distribution in visual recognition tasks. To address this problem, we first investigate the performance bottleneck of the two-stage learning framework via ablative study. Motivated by our discovery, we propose a unified distribution alignment strategy for long-tail visual recognition. Specifically, we develop an adaptive calibration function that enables us to adjust the classification scores for each data point. We then introduce a generalized re-weight method in the two-stage learning to balance the class prior, which provides a flexible and unified solution to diverse scenarios in visual recognition tasks. We validate our method by extensive experiments on four tasks, including image classification, semantic segmentation, object detection, and instance segmentation. Our approach achieves the state-of-the-art results across all four recognition tasks with a simple and unified framework.

We propose HOI Transformer to tackle human object interaction (HOI) detection in an end-to-end manner. Current approaches either decouple HOI task into separated stages of object detection and interaction classification or introduce surrogate interaction problem. In contrast, our method, named HOI Transformer, streamlines the HOI pipeline by eliminating the need for many hand-designed components. HOI Transformer reasons about the relations of objects and humans from global image context and directly predicts HOI instances in parallel. A quintuple matching loss is introduced to force HOI predictions in a unified way. Our method is conceptually much simpler and demonstrates improved accuracy. Without bells and whistles, HOI Transformer achieves 26.61% AP on HICO-DET and 52.9% AProleon V-COCO, surpassing previous methods with the advantage of being much simpler. We hope our approach will serve as a simple and effective alternative for HOI tasks. Code is available at https://github.com/bbepoch/HoiTransformer.

In this paper, we propose an efficient human pose estimation network (DANet) by learning deeply aggregated representations. Most existing models explore multi-scale infonnation mainly from features with different spatial sizes. Powerful multi-scale representations usually rely on the cascaded pyramid framework. This framework largely boosts the performance but in the meanwhile makes networks very deep and complex. Instead, we focus on exploiting multi-scale information from layers with different receptive-field sizes and then making full of use this infonnation by improving the fusion method. Specifically, we propose an orthogonal attention block (OAB) and a second-order fusion unit (SFU). The OAB learns multi-scale infonnation from different layers and enhances them by encouraging them to be diverse. The SFU adaptively selects and fuses diverse multi-scale infonnation and suppress the redundant ones. With the help of OAB and SFU, our networks could achieve comparable or even better accuracy with much smaller model complexity. Specifically, our DANet-72 achieves 71.0 in AP score on COCO val2017 with only 1.0G FLOPS. Its speed on a CPU platfonn achieves 58 Persons-Per-Second (PPS).

Knowledge distillation has shown great success in classification, however, it is still challenging for detection. In a typical image for detection, representations from different locations may have different contributions to detection targets, making the distillation hard to balance. In this paper, we propose a conditional distillation framework to distill the desired knowledge, namely knowledge that is beneficial in terms of both classification and localization for every instance. The framework introduces a learnable conditional decoding module, which retrieves information given each target instance as query. Specifically, we encode the condition information as query and use the teacher's representations as key. The attention between query and key is used to measure the contribution of different features, guided by a localization-recognition-sensitive auxiliary task. Extensive experiments demonstrate the efficacy of our method: we observe impressive improvements under various settings. Notably, we boost RetinaNet with ResNet-50 backbone from $37.4$ to $40.7$ mAP ($+3.3$) under $1\times$ schedule, that even surpasses the teacher ($40.4$ mAP) with ResNet-101 backbone under $3\times$ schedule. Code has been released on https://github.com/megvii-research/ICD.

Transformers, the de-facto standard for language modeling, have been recently applied for vision tasks. This paper introduces sparse queries for vision transformers to exploit the intrinsic spatial redundancy of natural images and save computational costs. Specifically, we propose a Dynamic Grained Encoder for vision transformers, which can adaptively assign a suitable number of queries to each spatial region. Thus it achieves a fine-grained representation in discriminative regions while keeping high efficiency. Besides, the dynamic grained encoder is compatible with most vision transformer frameworks. Without bells and whistles, our encoder allows the state-of-the-art vision transformers to reduce computational complexity by 40%-60% while maintaining comparable performance on image classification. Extensive experiments on object detection and segmentation further demonstrate the generalizability of our approach. Code is available at https://github.com/StevenGrove/vtpack.

In this paper, we comprehensively reveal the learning dynamics of normalized neural network using Stochastic Gradient Descent (with momentum) and Weight Decay (WD), named as Spherical Motion Dynamics (SMD). Most related works focus on studying behavior of effective learning rate" inequilibrium" state, i.e. assuming weight norm remains unchanged. However, their discussion on why this equilibrium can be reached is either absent or less convincing. Our work directly explores the cause of equilibrium, as a special state of SMD. Specifically, 1) we introduce the assumptions that can lead to equilibrium state in SMD, and prove equilibrium can be reached in a linear rate regime under given assumptions; 2) we propose ``angular update" as a substitute for effective learning rate to depict the state of SMD, and derive the theoretical value of angular update in equilibrium state; 3) we verify our assumptions and theoretical results on various large-scale computer vision tasks including ImageNet and MSCOCO with standard settings. Experiment results show our theoretical findings agree well with empirical observations. We also show that the behavior of angular update in SMD can produce interesting effect to the optimization of neural network in practice.

We present joint multi-dimension pruning (abbreviated as JointPruning), an effective method of pruning a network on three crucial aspects: spatial, depth and channel simultaneously. To tackle these three naturally different dimensions, we proposed a general framework by defining pruning as seeking the best pruning vector (i.e., the numerical value of layer-wise channel number, spatial size, depth) and construct a unique mapping from the pruning vector to the pruned network structures. Then we optimize the pruning vector with gradient update and model joint pruning as a numerical gradient optimization process. To overcome the challenge that there is no explicit function between the loss and the pruning vectors, we proposed self-adapted stochastic gradient estimation to construct a gradient path through network loss to pruning vectors and enable efficient gradient update. We show that the joint strategy discovers a better status than previous studies that focused on individual dimensions solely, as our method is optimized collaboratively across the three dimensions in a single end-to-end training and it is more efficient than the previous exhaustive methods. Extensive experiments on large-scale ImageNet dataset across a variety of network architectures MobileNet V1&V2&V3 and ResNet demonstrate the effectiveness of our proposed method. For instance, we achieve significant margins of 2.5% and 2.6% improvement over the state-of-the-art approach on the already compact MobileNet V1&V2 under an extremely large compression ratio.

We propose a simple yet effective proposal-based object detector, aiming at detecting highly-overlapped instances in crowded scenes. The key of our approach is to let each proposal predict a set of correlated instances rather than a single one in previous proposal-based frameworks. Equipped with new techniques such as EMD Loss and Set NMS, our detector can effectively handle the difficulty of detecting highly overlapped objects. On a FPN-Res50 baseline, our detector can obtain 4.9\% AP gains on challenging CrowdHuman dataset and 1.0\% $\text{MR}^{-2}$ improvements on CityPersons dataset, without bells and whistles. Moreover, on less crowed datasets like COCO, our approach can still achieve moderate improvement, suggesting the proposed method is robust to crowdedness.

In this work, we present a novel data-driven method for robust 6DoF object pose estimation from a single RGBD image. Unlike previous methods that directly regressing pose parameters, we tackle this challenging task with a keypoint-based approach. Specifically, we propose a deep Hough voting network to detect 3D keypoints of objects and then estimate the 6D pose parameters within a least-squares fitting manner. Our method is a natural extension of 2D-keypoint approaches that successfully work on RGB based 6DoF estimation. It allows us to fully utilize the geometric constraint of rigid objects with the extra depth information and is easy for a network to learn and optimize. Extensive experiments were conducted to demonstrate the effectiveness of 3D-keypoint detection in the 6D pose estimation task. Experimental results also show our method outperforms the state-of-the-art methods by large margins on several benchmarks. Code and video are available at https://github.com/ethnhe/PVN3D.git.

Recently, numerous handcrafted and searched networks have been applied for semantic segmentation. However, previous works intend to handle inputs with various scales in pre-defined static architectures, such as FCN, U-Net, and DeepLab series. This paper studies a conceptually new method to alleviate the scale variance in semantic representation, named dynamic routing. The proposed framework generates data-dependent routes, adapting to the scale distribution of each image. To this end, a differentiable gating function, called soft conditional gate, is proposed to select scale transform paths on the fly. In addition, the computational cost can be further reduced in an end-to-end manner by giving budget constraints to the gating function. We further relax the network level routing space to support multi-path propagations and skip-connections in each forward, bringing substantial network capacity. To demonstrate the superiority of the dynamic property, we compare with several static architectures, which can be modeled as special cases in the routing space. Extensive experiments are conducted on Cityscapes and PASCAL VOC 2012 to illustrate the effectiveness of the dynamic framework. Code is available at https://github.com/yanwei-li/DynamicRouting.

Traditional convolution-based generative adversarial networks synthesize images based on hierarchical local operations, where long-range dependency relation is implicitly modeled with a Markov chain. It is still not sufficient for categories with complicated structures. In this paper, we characterize long-range dependence with attentive normalization (AN), which is an extension to traditional instance normalization. Specifically, the input feature map is softly divided into several regions based on its internal semantic similarity, which are respectively normalized. It enhances consistency between distant regions with semantic correspondence. Compared with self-attention GAN, our attentive normalization does not need to measure the correlation of all locations, and thus can be directly applied to large-size feature maps without much computational burden. Extensive experiments on class-conditional image generation and semantic inpainting verify the efficacy of our proposed module.

Understanding interactions between humans and objects is one of the fundamental problems in visual classification and an essential step towards detailed scene understanding. Human-object interaction (HOI) detection strives to localize both the human and an object as well as the identification of complex interactions between them. Most existing HOI detection approaches are instance-centric where interactions between all possible human-object pairs are predicted based on appearance features and coarse spatial information. We argue that appearance features alone are insufficient to capture complex human-object interactions. In this paper, we therefore propose a novel fully-convolutional approach that directly detects the interactions between human-object pairs. Our network predicts interaction points, which directly localize and classify the inter-action. Paired with the densely predicted interaction vectors, the interactions are associated with human and object detections to obtain final predictions. To the best of our knowledge, we are the first to propose an approach where HOI detection is posed as a keypoint detection and grouping problem. Experiments are performed on two popular benchmarks: V-COCO and HICO-DET. Our approach sets a new state-of-the-art on both datasets. Code is available at https://github.com/vaesl/IP-Net.

Occluded person re-identification (ReID) aims to match occluded person images to holistic ones across dis-joint cameras. In this paper, we propose a novel framework by learning high-order relation and topology information for discriminative features and robust alignment. At first, we use a CNN backbone to learn feature maps and key-points estimation model to extract semantic local features. Even so, occluded images still suffer from occlusion and outliers. Then, we view the extracted local features of an image as nodes of a graph and propose an adaptive direction graph convolutional (ADGC) layer to pass relation information between nodes. The proposed ADGC layer can automatically suppress the message passing of meaningless features by dynamically learning direction and degree of linkage. When aligning two groups of local features, we view it as a graph matching problem and propose a cross-graph embedded-alignment (CGEA) layer to joint learn and embed topology information to local features, and straightly predict similarity score. The proposed CGEA layer can both take full use of alignment learned by graph matching and replace sensitive one-to-one alignment with a robust soft one. Finally, extensive experiments on occluded, partial, and holistic ReID tasks show the effectiveness of our proposed method. Specifically, our framework significantly outperforms state-of-the-art by $6.5\%$ mAP scores on Occluded-Duke dataset.

The Learnable Tree Filter presents a remarkable approach to model structure-preserving relations for semantic segmentation. Nevertheless, the intrinsic geometric constraint forces it to focus on the regions with close spatial distance, hindering the effective long-range interactions. To relax the geometric constraint, we give the analysis by reformulating it as a Markov Random Field and introduce a learnable unary term. Besides, we propose a learnable spanning tree algorithm to replace the original non-differentiable one, which further improves the flexibility and robustness. With the above improvements, our method can better capture long range dependencies and preserve structural details with linear complexity, which is extended to several vision tasks for more generic feature transform. Extensive experiments on object detection/instance segmentation demonstrate the consistent improvements over the original version. For semantic segmentation, we achieve leading performance (82.1% mIoU) on the Cityscapes benchmark without bells-and whistles. Code is available at https://github.com/StevenGrove/LearnableTreeFilterV2.

The Feature Pyramid Network (FPN) presents a remarkable approach to alleviate the scale variance in object representation by performing instance-level assignments. Nevertheless, this strategy ignores the distinct characteristics of different sub-regions in an instance. To this end, we propose a fine-grained dynamic head to conditionally select a pixel-level combination of FPN features from different scales for each instance, which further releases the ability of multi-scale feature representation. Moreover, we design a spatial gate with the new activation function to reduce computational complexity dramatically through spatially sparse convolutions. Extensive experiments demonstrate the effectiveness and efficiency of the proposed method on several state-of-the-art detection benchmarks. Code is available at https://github.com/StevenGrove/DynamicHead.

Recent research on super-resolution has achieved great success due to the development of deep convolutional neural networks (DCNNs). However, super-resolution of arbitrary scale factor has been ignored for a long time. Most previous researchers regard super-resolution of differentscale factors as independent tasks. They train a specific model for each scale factor which is inefficient in computing, and prior work only take the super-resolution of several integer scale factors into consideration. In this work,we propose a novel method called Meta-SR to firstly solve super-resolution of arbitrary scale factor (including non-integer scale factors) with a single model. In our Meta-SR,the Meta-Upscale Module is proposed to replace the traditional upscale module. For arbitrary scale factor, the Meta-Upscale Module dynamically predicts the weights of the up-scale filters by taking the scale factor as input and use these weights to generate the HR image of arbitrary size. For any low-resolution image, our Meta-SR can continuously zoomin it with arbitrary scale factor by only using a single model.We evaluated the proposed method through extensive experiments on widely used benchmark datasets on single image super-resolution. The experimental results show the superiority of our Meta-Upscale.

This paper introduces an extremely efficient CNN architecture named DFANet for semantic segmentation under resource constraints. Our proposed network starts from a single lightweight backbone and aggregates discriminative features through sub-network and sub-stage cascade respectively. Based on the multi-scale feature propagation, DFANet substantially reduces the number of parameters, but still obtains sufficient receptive field and enhances the model learning ability, which strikes a balance between the speed and segmentation performance. Experiments on Cityscapes and CamVid datasets demonstrate the superior performance of DFANet with 8$\times$ less FLOPs and 2$\times$ faster than the existing state-of-the-art real-time semantic segmentation methods while providing comparable accuracy. Specifically, it achieves 70.3\% Mean IOU on the Cityscapes test dataset with only 1.7 GFLOPs and a speed of 160 FPS on one NVIDIA Titan X card, and 71.3\% Mean IOU with 3.4 GFLOPs while inferring on a higher resolution image.

This paper considers a realistic problem in person re-identification (re-ID) task, i.e., partial re-ID. Under partial re-ID scenario, the images may contain a partial observation of a pedestrian. If we directly compare a partial pedestrian image with a holistic one, the extreme spatial misalignment significantly compromises the discriminative ability of the learned representation. We propose a Visibility-aware Part Model (VPM) for partial re-ID, which learns to perceive the visibility of regions through self-supervision. The visibility awareness allows VPM to extract region-level features and compare two images with focus on their shared regions (which are visible on both images). VPM gains two-fold benefit toward higher accuracy for partial re-ID. On the one hand, compared with learning a global feature, VPM learns region-level features and thus benefits from fine-grained information. On the other hand, with visibility awareness, VPM is capable to estimate the shared regions between two images and thus suppresses the spatial misalignment. Experimental results confirm that our method significantly improves the learned feature representation and the achieved accuracy is on par with the state of the art.

In this paper, we introduce a new large-scale object detection dataset, Objects365, which has 365 object categories over 600K training images. More than 10 million, high-quality bounding boxes are manually labeled through a three-step, carefully designed annotation pipeline. It is the largest object detection dataset (with full annotation) so far and establishes a more challenging benchmark for the community. Objects365 can serve as a better feature learning dataset for localization-sensitive tasks like object detection and semantic segmentation. The Objects365 pre-trained models significantly outperform ImageNet pre-trained models with 5.6 points gain (42 vs 36.4) based on the standard setting of 90K iterations on COCO benchmark. Even compared with much long training time like 540K iterations, our Objects365 pretrained model with 90K iterations still have 2.7 points gain (42 vs 39.3). Meanwhile, the finetuning time can be greatly reduced (up to 10 times) when reaching the same accuracy. Better generalization ability of Object365 has also been verified on CityPersons, VOC segmentation, and ADE tasks. The dataset as well as the pretrained-models have been released at www.objects365.org.

Most previous image matting methods require a roughly-specificed trimap as input, and estimate fractional alpha values for all pixels that are in the unknown region of the trimap. In this paper, we argue that directly estimating the alpha matte from a coarse trimap is a major limitation of previous methods, as this practice tries to address two difficult and inherently different problems at the same time: identifying true blending pixels inside the trimap region, and estimate accurate alpha values for them. We propose AdaMatting, a new end-to-end matting framework that disentangles this problem into two sub-tasks: trimap adaptation and alpha estimation. Trimap adaptation is a pixel-wise classification problem that infers the global structure of the input image by identifying definite foreground, background, and semi-transparent image regions. Alpha estimation is a regression problem that calculates the opacity value of each blended pixel. Our method separately handles these two sub-tasks within a single deep convolutional neural network (CNN). Extensive experiments show that AdaMatting has additional structure awareness and trimap fault-tolerance. Our method achieves the state-of-the-art performance on Adobe Composition-1k dataset both qualitatively and quantitatively. It is also the current best-performing method on the alphamatting.com online evaluation for all commonly-used metrics.

In this paper, we propose a novel meta learning approach for automatic channel pruning of very deep neural networks. We first train a PruningNet, a kind of meta network, which is able to generate weight parameters for any pruned structure given the target network. We use a simple stochastic structure sampling method for training the PruningNet. Then, we apply an evolutionary procedure to search for good-performing pruned networks. The search is highly efficient because the weights are directly generated by the trained PruningNet and we do not need any finetuning at search time. With a single PruningNet trained for the target network, we can search for various Pruned Networks under different constraints with little human participation. Compared to the state-of-the-art pruning methods, we have demonstrated superior performances on MobileNet V1/V2 and ResNet. Codes are available on https://github.com/liuzechun/MetaPruning.

Real-time generic object detection on mobile platforms is a crucial but challenging computer vision task. Prior lightweight CNN-based detectors are inclined to use one-stage pipeline. In this paper, we investigate the effectiveness of two-stage detectors in real-time generic detection and propose a lightweight two-stage detector named ThunderNet. In the backbone part, we analyze the drawbacks in previous lightweight backbones and present a lightweight backbone designed for object detection. In the detection part, we exploit an extremely efficient RPN and detection head design. To generate more discriminative feature representation, we design two efficient architecture blocks, Context Enhancement Module and Spatial Attention Module. At last, we investigate the balance between the input resolution, the backbone, and the detection head. Benefit from the highly efficient backbone and detection part design, ThunderNet surpasses previous lightweight one-stage detectors with only 40% of the computational cost on PASCAL VOC and COCO benchmarks. Without bells and whistles, ThunderNet runs at 24.1 fps on an ARM-based device with 19.2 AP on COCO. To the best of our knowledge, this is the first real-time detector reported on ARM platforms. Code will be released for paper reproduction.

Object detectors are usually equipped with backbone networks designed for image classification. It might be sub-optimal because of the gap between the tasks of image classification and object detection. In this work, we present DetNAS to use Neural Architecture Search (NAS) for the design of better backbones for object detection. It is non-trivial because detection training typically needs ImageNetpre-training while NAS systems require accuracies on the target detection task as supervisory signals. Based on the technique of one-shot supernet, which contains all possible networks in the search space, we propose a framework for backbone search on object detection. We train the supernet under the typical detector training schedule: ImageNet pre-training and detection fine-tuning. Then, the architecture search is performed on the trained supernet, using the detection task as the guidance. This framework makes NAS on backbones very efficient. In experiments, we show the effectiveness of DetNAS on various detectors, for instance, one-stage RetinaNetand the two-stage FPN. We empirically find that networks searched on object detection shows consistent superiority compared to those searched on ImageNet classification. The resulting architecture achieves superior performance than hand-crafted networks on COCO with much less FLOPs complexity.

Learning discriminative global features plays a vital role in semantic segmentation. And most of the existing methods adopt stacks of local convolutions or non-local blocks to capture long-range context. However, due to the absence of spatial structure preservation, these operators ignore the object details when enlarging receptive fields. In this paper, we propose the learnable tree filter to form a generic tree filtering module that leverages the structural property of minimal spanning tree to model long-range dependencies while preserving the details. Furthermore, we propose a highly efficient linear-time algorithm to reduce resource consumption. Thus, the designed modules can be plugged into existing deep neural networks conveniently. To this end, tree filtering modules are embedded to formulate a unified framework for semantic segmentation. We conduct extensive ablation studies to elaborate on the effectiveness and efficiency of the proposed method. Specifically, it attains better performance with much less overhead compared with the classic PSP block and Non-local operation under the same backbone. Our approach is proved to achieve consistent improvements on several benchmarks without bells-and-whistles. Code and models are available at https://github.com/StevenGrove/TreeFilter-Torch.

We study the problem of grounding distributional representations of texts on the visual domain, namely visual-semantic embeddings (VSE for short). Begin with an insightful adversarial attack on VSE embeddings, we show the limitation of current frameworks and image-text datasets (e.g., MS-COCO) both quantitatively and qualitatively. The large gap between the number of possible constitutions of real-world semantics and the size of parallel data, to a large extent, restricts the model to establish a strong link between textual semantics and visual concepts. We alleviate this problem by augmenting the MS-COCO image captioning datasets with textual contrastive adversarial samples. These samples are synthesized using language priors of human and the WordNet knowledge base, and enforce the model to ground learned embeddings to concrete concepts within the image. This simple but powerful technique brings a noticeable improvement over the baselines on a diverse set of downstream tasks, in addition to defending known-type adversarial attacks. Codes are available at https://github.com/ExplorerFreda/VSE-C.

The topic of multi-person pose estimation has been largely improved recently, especially with the development of convolutional neural network. However, there still exist a lot of challenging cases, such as occluded keypoints, invisible keypoints and complex background, which cannot be well addressed. In this paper, we present a novel network structure called Cascaded Pyramid Network (CPN) which targets to relieve the problem from these "hard" keypoints. More specifically, our algorithm includes two stages: GlobalNet and RefineNet. GlobalNet is a feature pyramid network which can successfully localize the "simple" keypoints like eyes and hands but may fail to precisely recognize the occluded or invisible keypoints. Our RefineNet tries explicitly handling the "hard" keypoints by integrating all levels of feature representations from the GlobalNet together with an online hard keypoint mining loss. In general, to address the multi-person pose estimation problem, a top-down pipeline is adopted to first generate a set of human bounding boxes based on a detector, followed by our CPN for keypoint localization in each human bounding box. Based on the proposed algorithm, we achieve state-of-art results on the COCO keypoint benchmark, with average precision at 73.0 on the COCO test-dev dataset and 72.1 on the COCO test-challenge dataset, which is a 19% relative improvement compared with 60.5 from the COCO 2016 keypoint challenge. Code1 and the detection results for person used will be publicly available for further research.

The development of object detection in the era of deep learning, from R-CNN [11], Fast/Faster R-CNN [10, 31] to recent Mask R-CNN [14] and RetinaNet [24], mainly come from novel network, new framework, or loss design. However, mini-batch size, a key factor for the training of deep neural networks, has not been well studied for object detection. In this paper, we propose a Large Mini-Batch Object Detector (MegDet) to enable the training with a large mini-batch size up to 256, so that we can effectively utilize at most 128 GPUs to significantly shorten the training time. Technically, we suggest a warmup learning rate policy and Cross-GPU Batch Normalization, which together allow us to successfully train a large mini-batch detector in much less time (e.g., from 33 hours to 4 hours), and achieve even better accuracy. The MegDet is the backbone of our submission (mmAP 52.5%) to COCO 2017 Challenge, where we won the 1st place of Detection task.

Detecting individual pedestrians in a crowd remains a challenging problem since the pedestrians often gather together and occlude each other in real-world scenarios. In this paper, we first explore how a state-of-the-art pedestrian detector is harmed by crowd occlusion via experimentation, providing insights into the crowd occlusion problem. Then, we propose a novel bounding box regression loss specifically designed for crowd scenes, termed repulsion loss. This loss is driven by two motivations: the attraction by target, and the repulsion by other surrounding objects. The repulsion term prevents the proposal from shifting to surrounding objects thus leading to more crowd-robust localization. Our detector trained by repulsion loss outperforms the state-of-the-art methods with a significant improvement in occlusion cases.

We introduce an extremely computation-efficient CNN architecture named ShuffleNet, which is designed specially for mobile devices with very limited computing power (e.g., 10-150 MFLOPs). The new architecture utilizes two new operations, pointwise group convolution and channel shuffle, to greatly reduce computation cost while maintaining accuracy. Experiments on ImageNet classification and MS COCO object detection demonstrate the superior performance of ShuffleNet over other structures, e.g. lower top-1 error (absolute 7.8%) than recent MobileNet [12] on ImageNet classification task, under the computation budget of 40 MFLOPs. On an ARM-based mobile device, ShuffleNet achieves ~13× actual speedup over AlexNet while maintaining comparable accuracy.

We propose a novel and flexible anchor mechanism named MetaAnchor for object detection frameworks. Unlike many previous detectors model anchors via a predefined manner, in MetaAnchor anchor functions could be dynamically generated from the arbitrary customized prior boxes. Taking advantage of weight prediction, MetaAnchor is able to work with most of the anchor-based object detection systems such as RetinaNet. Compared with the predefined anchor scheme, we empirically find that MetaAnchor is more robust to anchor settings and bounding box distributions; in addition, it also shows the potential on the transfer task. Our experiment on COCO detection task shows MetaAnchor consistently outperforms the counterparts in various scenarios.

In content-based image retrieval, SIFT feature and the feature from deep convolutional neural network (CNN) have demonstrated promising performance. To fully explore both visual features in a unified framework for effective and efficient retrieval, we propose a collaborative index embedding method to implicitly integrate the index matrices of them. We formulate the index embedding as an optimization problem from the perspective of neighborhood sharing and solve it with an alternating index update scheme. After the iterative embedding, only the embedded CNN index is kept for on-line query, which demonstrates significant gain in retrieval accuracy, with very economical memory cost. Extensive experiments have been conducted on the public datasets with million-scale distractor images. The experimental results reveal that, compared with the recent state-of-the-art retrieval algorithms, our approach achieves competitive accuracy performance with less memory overhead and efficient query computation.

The problem of quantizing the activations of a deep neural network is considered. An examination of the popular binary quantization approach shows that this consists of approximating a classical non-linearity, the hyperbolic tangent, by two functions: a piecewise constant sign function, which is used in feedforward network computations, and a piecewise linear hard tanh function, used in the backpropagation step during network learning. The problem of approximating the widely used ReLU non-linearity is then considered. An half-wave Gaussian quantizer (HWGQ) is proposed for forward approximation and shown to have efficient implementation, by exploiting the statistics of of network activations and batch normalization operations. To overcome the problem of gradient mismatch, due to the use of different forward and backward approximations, several piece-wise backward approximators are then investigated. The implementation of the resulting quantized network, denoted as HWGQ-Net, is shown to achieve much closer performance to full precision networks, such as AlexNet, ResNet, GoogLeNet and VGG-Net, than previously available low-precision networks, with 1-bit binary weights and 2-bit quantized activations.

One of recent trends [31, 32, 14] in network architecture design is stacking small filters (e.g., 1×1 or 3×3) in the entire network because the stacked small filters is more efficient than a large kernel, given the same computational complexity. However, in the field of semantic segmentation, where we need to perform dense per-pixel prediction, we find that the large kernel (and effective receptive field) plays an important role when we have to perform the classification and localization tasks simultaneously. Following our design principle, we propose a Global Convolutional Network to address both the classification and localization issues for the semantic segmentation. We also suggest a residual-based boundary refinement to further refine the object boundaries. Our approach achieves state-of-art performance on two public benchmarks and significantly outperforms previous results, 82.2% (vs 80.2%) on PASCAL VOC 2012 dataset and 76.9% (vs 71.8%) on Cityscapes dataset.

In this paper, we introduce a new channel pruning method to accelerate very deep convolutional neural networks. Given a trained CNN model, we propose an iterative two-step algorithm to effectively prune each layer, by a LASSO regression based channel selection and least square reconstruction. We further generalize this algorithm to multi-layer and multi-branch cases. Our method reduces the accumulated error and enhance the compatibility with various architectures. Our pruned VGG-16 achieves the state-of-the-art results by 5× speed-up along with only 0.3% increase of error. More importantly, our method is able to accelerate modern networks like ResNet, Xception and suffers only 1.4%, 1.0% accuracy loss under 2× speedup respectively, which is significant.

This paper revisits the classical Bayesian face recognition algorithm from Baback Moghaddam et al. and proposes enhancements tailored to face verification, the problem of predicting whether or not a pair of facial images share the same identity. Like a variety of face verification algorithms, the original Bayesian face model only considers the appearance difference between two faces rather than the raw images themselves. However, we argue that such a fixed and blind projection may prematurely reduce the separability between classes. Consequently, we model two facial images jointly with an appropriate prior that considers intra- and extra-personal variations over the image pairs. This joint formulation is trained using a principled EM algorithm, while testing involves only efficient closed-formed computations that are suitable for real-time practical deployment. Supporting theoretical analyses investigate computational complexity, scale-invariance properties, and convergence issues. We also detail important relationships with existing algorithms, such as probabilistic linear discriminant analysis and metric learning. Finally, on extensive experimental evaluations, the proposed model is superior to the classical Bayesian face algorithm and many alternative state-of-the-art supervised approaches, achieving the best test accuracy on three challenging datasets, Labeled Face in Wild, Multi-PIE, and YouTube Faces, all with unparalleled computational efficiency.

State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet [1] and Fast R-CNN [2] have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features-using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model [3] , our detection system has a frame rate of 5 fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks. Code has been made publicly available.

Most object detectors contain two important components: a feature extractor and an object classifier. The feature extractor has rapidly evolved with significant research efforts leading to better deep convolutional architectures. The object classifier, however, has not received much attention and many recent systems (like SPPnet and Fast/Faster R-CNN) use simple multi-layer perceptrons. This paper demonstrates that carefully designing deep networks for object classification is just as important. We experiment with region-wise classifier networks that use shared, region-independent convolutional features. We call them "Networks on Convolutional feature maps" (NoCs). We discover that aside from deep feature maps, a deep and convolutional per-region classifier is of particular importance for object detection, whereas latest superior image classification models (such as ResNets and GoogLeNets) do not directly lead to good detection accuracy without using such a per-region classifier. We show by experiments that despite the effective ResNets and Faster R-CNN systems, the design of NoCs is an essential element for the 1st-place winning entries in ImageNet and MS COCO challenges 2015.

Semantic segmentation research has recently witnessed rapid progress, but many leading methods are unable to identify object instances. In this paper, we present Multitask Network Cascades for instance-aware semantic segmentation. Our model consists of three networks, respectively differentiating instances, estimating masks, and categorizing objects. These networks form a cascaded structure, and are designed to share their convolutional features. We develop an algorithm for the nontrivial end-to-end training of this causal, cascaded structure. Our solution is a clean, single-step training framework and can be generalized to cascades that have more stages. We demonstrate state-of-the-art instance-aware semantic segmentation accuracy on PASCAL VOC. Meanwhile, our method takes only 360ms testing an image using VGG-16, which is two orders of magnitude faster than previous systems for this challenging problem. As a by product, our method also achieves compelling object detection results which surpass the competitive Fast/Faster R-CNN systems. The method described in this paper is the foundation of our submissions to the MS COCO 2015 segmentation competition, where we won the 1st place.

Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers - 8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Large-scale data is of crucial importance for learning semantic segmentation models, but annotating per-pixel masks is a tedious and inefficient procedure. We note that for the topic of interactive image segmentation, scribbles are very widely used in academic research and commercial software, and are recognized as one of the most userfriendly ways of interacting. In this paper, we propose to use scribbles to annotate images, and develop an algorithm to train convolutional networks for semantic segmentation supervised by scribbles. Our algorithm is based on a graphical model that jointly propagates information from scribbles to unmarked pixels and learns network parameters. We present competitive object semantic segmentation results on the PASCAL VOC dataset by using scribbles as annotations. Scribbles are also favored for annotating stuff (e.g., water, sky, grass) that has no well-defined shape, and our method shows excellent results on the PASCALCONTEXT dataset thanks to extra inexpensive scribble annotations. Our scribble annotations on PASCAL VOC are available at http://research.microsoft.com/en-us/um/ people/jifdai/downloads/scribble_sup.

We present region-based, fully convolutional networks for accurate and efficient object detection. In contrast to previous region-based detectors such as Fast/Faster R-CNN that apply a costly per-region subnetwork hundreds of times, our region-based detector is fully convolutional with almost all computation shared on the entire image. To achieve this goal, we propose position-sensitive score maps to address a dilemma between translation-invariance in image classification and translation-variance in object detection. Our method can thus naturally adopt fully convolutional image classifier backbones, such as the latest Residual Networks (ResNets), for object detection. We show competitive results on the PASCAL VOC datasets (e.g., 83.6% mAP on the 2007 set) with the 101-layer ResNet. Meanwhile, our result is achieved at a test-time speed of 170ms per image, 2.5-20 times faster than the Faster R-CNN counterpart. Code is made publicly available at: https://github.com/daijifeng001/r-fcn.

This paper aims to accelerate the test-time computation of convolutional neural networks (CNNs), especially very deep CNNs [1] that have substantially impacted the computer vision community. Unlike previous methods that are designed for approximating linear filters or linear responses, our method takes the nonlinear units into account. We develop an effective solution to the resulting nonlinear optimization problem without the need of stochastic gradient descent (SGD). More importantly, while previous methods mainly focus on optimizing one or two layers, our nonlinear method enables an asymmetric reconstruction that reduces the rapidly accumulated error when multiple (e.g., ≥ 10) layers are approximated. For the widely used very deep VGG-16 model [1] , our method achieves a whole-model speedup of 4 × with merely a 0.3 percent increase of top-5 error in ImageNet classification. Our 4 × accelerated VGG-16 model also shows a graceful accuracy degradation for object detection when plugged into the Fast R-CNN detector [2] .

This paper presents a highly efficient and accurate regression approach for face alignment. Our approach has two novel components: 1) a set of local binary features and 2) a locality principle for learning those features. The locality principle guides us to learn a set of highly discriminative local binary features for each facial landmark independently. The obtained local binary features are used to jointly learn a linear regression for the final output. This approach achieves the state-of-the-art results when tested on the most challenging benchmarks to date. Furthermore, because extracting and regressing local binary features are computationally very cheap, our system is much faster than previous methods. It achieves over 3000 frames per second (FPS) on a desktop or 300 FPS on a mobile phone for locating a few dozens of landmarks. We also study a key issue that is important but has received little attention in the previous research, which is the face detector used to initialize alignment. We investigate several face detectors and perform quantitative evaluation on how they affect alignment accuracy. We find that an alignment friendly detector can further greatly boost the accuracy of our alignment method, reducing the error up to 16% relatively. To facilitate practical usage of face detection/alignment methods, we also propose a convenient metric to measure how good a detector is for alignment initialization.

Bezigons, i.e., closed paths composed of Bézier curves, have been widely employed to describe shapes in image vectorization results. However, most existing vectorization techniques infer the bezigons by simply approximating an intermediate vector representation (such as polygons). Consequently, the resultant bezigons are sometimes imperfect due to accumulated errors, fitting ambiguities, and a lack of curve priors, especially for low-resolution images. In this paper, we describe a novel method for vectorizing clipart images. In contrast to previous methods, we directly optimize the bezigons rather than using other intermediate representations; therefore, the resultant bezigons are not only of higher fidelity compared with the original raster image but also more reasonable because they were traced by a proficient expert. To enable such optimization, we have overcome several challenges and have devised a differentiable data energy as well as several curve-based prior terms. To improve the efficiency of the optimization, we also take advantage of the local control property of bezigons and adopt an overlapped piecewise optimization strategy. The experimental results show that our method outperforms both the current state-of-the-art method and commonly used commercial software in terms of bezigon quality.

We extends the previous 2D cascaded object pose regression work [9] in two aspects so that it works better for 3D articulated objects. Our first contribution is 3D pose-indexed features that generalize the previous 2D parameterized features and achieve better invariance to 3D transformations. Our second contribution is a principled hierarchical regression that is adapted to the articulated object structure. It is therefore more accurate and faster. Comprehensive experiments verify the state-of-the-art accuracy and efficiency of the proposed approach on the challenging 3D hand pose estimation problem, on a public dataset and our new dataset.

The topic of semantic segmentation has witnessed considerable progress due to the powerful features learned by convolutional neural networks (CNNs) [13]. The current leading approaches for semantic segmentation exploit shape information by extracting CNN features from masked image regions. This strategy introduces artificial boundaries on the images and may impact the quality of the extracted features. Besides, the operations on the raw image domain require to compute thousands of networks on a single image, which is time-consuming. In this paper, we propose to exploit shape information via masking convolutional features. The proposal segments (e.g., super-pixels) are treated as masks on the convolutional feature maps. The CNN features of segments are directly masked out from these maps and used to train classifiers for recognition. We further propose a joint method to handle objects and “stuff” (e.g., grass, sky, water) in the same framework. State-of-the-art results are demonstrated on benchmarks of PASCAL VOC and new PASCAL-CONTEXT, with a compelling computational speed.

Though recent advanced convolutional neural networks (CNNs) have been improving the image recognition accuracy, the models are getting more complex and time-consuming. For real-world applications in industrial and commercial scenarios, engineers and developers are often faced with the requirement of constrained time budget. In this paper, we investigate the accuracy of CNNs under constrained time cost. Under this constraint, the designs of the network architectures should exhibit as trade-offs among the factors like depth, numbers of filters, filter sizes, etc. With a series of controlled comparisons, we progressively modify a baseline model while preserving its time complexity. This is also helpful for understanding the importance of the factors in network designs. We present an architecture that achieves very competitive accuracy in the ImageNet dataset (11.8% top-5 error, 10-view test), yet is 20% faster than “AlexNet” [14] (16.0% top-5 error, 10-view test).

The manipulation of panoramic/wide-angle images is usually achieved via image warping. Though various techniques have been developed for preserving shapes and straight lines for warping, these are not sufficient for panoramic/wide-angle images. The image projections will turn the straight lines into curved “geodesic lines”, and it is fundamentally impossible to keep all these lines straight. In this work, we propose a geodesic-preserving method for content-aware image warping. An energy term is introduced to preserve the geodesic appearance of the geodesic lines, and can be used with shape-preserving terms. Our method is demonstrated in various applications, including rectangling panoramas, resizing panoramic/wide-angle images, and wide-angle image manipulation. An extension to ellipse preservation for general images is also presented.

Random forest is well known as one of the best learning methods. In spite of its great success, it also has certain drawbacks: the heuristic learning rule does not effectively minimize the global training loss; the model size is usually too large for many real applications. To address the issues, we propose two techniques, global refinement and global pruning, to improve a pre-trained random forest. The proposed global refinement jointly relearns the leaf nodes of all trees under a global objective function so that the complementary information between multiple trees is well exploited. In this way, the fitting power of the forest is significantly enhanced. The global pruning is developed to reduce the model size as well as the over-fitting risk. The refined model has better performance and smaller storage cost, as verified in extensive experiments.

This paper addresses the problem of learning long binary codes from high-dimensional data. We observe that two key challenges arise while learning and using long binary codes: (1) lack of an effective regularizer for the learned high-dimensional mapping and (2) high computational cost for computing long codes. In this paper, we overcome both these problems by introducing a sparsity encouraging regularizer that reduces the effective number of parameters involved in the learned projection operator. This regularizer not only reduces overfitting but, due to the sparse nature of the projection matrix, also leads to a dramatic reduction in the computational cost. To evaluate the effectiveness of our method, we analyze its performance on the problems of nearest neighbour search, image retrieval and image classification. Experiments on a number of challenging datasets show that our method leads to better accuracy than dense projections (ITQ [11] and LSH [16]) with the same code lengths, and meanwhile is over an order of magnitude faster. Furthermore, our method is also more accurate and faster than other recently proposed methods for speeding up high-dimensional binary encoding.

This paper aims to accelerate the test-time computation of deep convolutional neural networks (CNNs). Unlike existing methods that are designed for approximating linear filters or linear responses, our method takes the nonlinear units into account. We minimize the reconstruction error of the nonlinear responses, subject to a low-rank constraint which helps to reduce the complexity of filters. We develop an effective solution to this constrained nonlinear optimization problem. An algorithm is also presented for reducing the accumulated error when multiple layers are approximated. A whole-model speedup ratio of 4× is demonstrated on a large network trained for ImageNet, while the top-5 error rate is only increased by 0.9%. Our accelerated model has a comparably fast speed as the “AlexNet” [11], but is 4.7% more accurate.

Recent leading approaches to semantic segmentation rely on deep convolutional networks trained with human-annotated, pixel-level segmentation masks. Such pixel-accurate supervision demands expensive labeling effort and limits the performance of deep networks that usually benefit from more training data. In this paper, we propose a method that achieves competitive accuracy but only requires easily obtained bounding box annotations. The basic idea is to iterate between automatically generating region proposals and training convolutional networks. These two steps gradually recover segmentation masks for improving the networks, and vise versa. Our method, called "BoxSup", produces competitive results (e.g., 62.0% mAP for validation) supervised by boxes only, on par with strong baselines (e.g., 63.8% mAP) fully supervised by masks under the same setting. By leveraging a large amount of bounding boxes, BoxSup further yields state-of-the-art results on PASCAL VOC 2012 and PASCAL-CONTEXT [26].

Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset.

To break down the geometry assumptions of traditional motion models (e.g., homography, affine), warping-based motion model recently becomes very popular and is adopted in many latest applications (e.g., image stitching, video stabilization). With high degrees of freedom, the accuracy of model heavily relies on data-terms (keypoint correspondences). In some low-texture environments (e.g., indoor) where keypoint feature is insufficient or unreliable, the warping model is often erroneously estimated. In this paper we propose a simple and effective approach by considering both keypoint and line segment correspondences as data-term. Line segment is a prominent feature in artificial environments and it can supply sufficient geometrical and structural information of scenes, which not only helps guild to a correct warp in low-texture condition, but also prevents the undesired distortion induced by warping. The combination aims to complement each other and benefit for a wider range of scenes. Our method is general and can be ported to many existing applications. Experiments demonstrate that using dual-feature yields more robust and accurate result especially for those low-texture images.

We study the problem of automatically removing outliers from noisy data, with application for removing outlier images from an image collection. We address this problem by utilizing the reconstruction errors of an autoencoder. We observe that when data are reconstructed from low-dimensional representations, the inliers and the outliers can be well separated according to their reconstruction errors. Based on this basic observation, we gradually inject discriminative information in the learning process of an autoencoder to make the inliers and the outliers more separable. Experiments on a variety of image datasets validate our approach.

State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully-convolutional network that simultaneously predicts object bounds and objectness scores at each position. RPNs are trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. With a simple alternating optimization, RPN and Fast R-CNN can be trained to share convolutional features. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007 (73.2% mAP) and 2012 (70.4% mAP) using 300 proposals per image. Code is available at https://github.com/ShaoqingRen/faster_rcnn.

Existing deep convolutional neural networks (CNNs) require a fixed-size (e.g., 224 × 224) input image. This requirement is "artificial" and may reduce the recognition accuracy for the images or sub-images of an arbitrary size/scale. In this work, we equip the networks with another pooling strategy, "spatial pyramid pooling", to eliminate the above requirement. The new network structure, called SPP-net, can generate a fixed-length representation regardless of image size/scale. Pyramid pooling is also robust to object deformations. With these advantages, SPP-net should in general improve all CNN-based image classification methods. On the ImageNet 2012 dataset, we demonstrate that SPP-net boosts the accuracy of a variety of CNN architectures despite their different designs. On the Pascal VOC 2007 and Caltech101 datasets, SPP-net achieves state-of-the-art classification results using a single full-image representation and no fine-tuning. The power of SPP-net is also significant in object detection. Using SPP-net, we compute the feature maps from the entire image only once, and then pool features in arbitrary regions (sub-images) to generate fixed-length representations for training the detectors. This method avoids repeatedly computing the convolutional features. In processing test images, our method is 24-102 × faster than the R-CNN method, while achieving better or comparable accuracy on Pascal VOC 2007. In ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2014, our methods rank #2 in object detection and #3 in image classification among all 38 teams. This manuscript also introduces the improvement made for this competition.

In this paper, we propose a novel model, a discriminatively learned iterative shrinkage (DLIS) model, for color image denoising. The DLIS is a generalization of wavelet shrinkage by iteratively performing shrinkage over patch groups and whole image aggregation. We discriminatively learn the shrinkage functions and basis from the training pairs of noisy/noise-free images, which can adaptively handle different noise characteristics in luminance/chrominance channels, and the unknown structured noise in real-captured color images. Furthermore, to remove the splotchy real color noises, we design a Laplacian pyramid-based denoising framework to progressively recover the clean image from the coarsest scale to the finest scale by the DLIS model learned from the real color noises. Experiments show that our proposed approach can achieve the state-of-the-art denoising results on both synthetic denoising benchmark and real-captured color images.

Sparse coding is a widely involved technique in computer vision. However, the expensive computational cost can hamper its applications, typically when the codebook size must be limited due to concerns on running time. In this paper, we study a special case of sparse coding in which the codebook is a Cartesian product of two subcodebooks. We present algorithms to decompose this sparse coding problem into smaller subproblems, which can be separately solved. Our solution, named as Product Sparse Coding (PSC), reduces the time complexity from O(K) to O(√K) in the codebook size K. In practice, this can be 20-100× faster than standard sparse coding. In experiments we demonstrate the efficiency and quality of this method on the applications of image classification and image retrieval.

We propose a novel motion model, SteadyFlow, to represent the motion between neighboring video frames for stabilization. A SteadyFlow is a specific optical flow by enforcing strong spatial coherence, such that smoothing feature trajectories can be replaced by smoothing pixel profiles, which are motion vectors collected at the same pixel location in the SteadyFlow over time. In this way, we can avoid brittle feature tracking in a video stabilization system. Besides, SteadyFlow is a more general 2D motion model which can deal with spatially-variant motion. We initialize the SteadyFlow by optical flow and then discard discontinuous motions by a spatial-temporal analysis and fill in missing regions by motion completion. Our experiments demonstrate the effectiveness of our stabilization on real-world challenging videos.

We present a realtime hand tracking system using a depth sensor. It tracks a fully articulated hand under large viewpoints in realtime (25 FPS on a desktop without using a GPU) and with high accuracy (error below 10 mm). To our knowledge, it is the first system that achieves such robustness, accuracy, and speed simultaneously, as verified on challenging real data. Our system is made of several novel techniques. We model a hand simply using a number of spheres and define a fast cost function. Those are critical for realtime performance. We propose a hybrid method that combines gradient based and stochastic optimization methods to achieve fast convergence and good accuracy. We present new finger detection and hand initialization methods that greatly enhance the robustness of tracking.

This paper presents a highly efficient, very accurate regression approach for face alignment. Our approach has two novel components: a set of local binary features, and a locality principle for learning those features. The locality principle guides us to learn a set of highly discriminative local binary features for each facial landmark independently. The obtained local binary features are used to jointly learn a linear regression for the final output. Our approach achieves the state-of-the-art results when tested on the current most challenging benchmarks. Furthermore, because extracting and regressing local binary features is computationally very cheap, our system is much faster than previous methods. It achieves over 3, 000 fps on a desktop or 300 fps on a mobile phone for locating a few dozens of landmarks.

Recent progresses in salient object detection have exploited the boundary prior, or background information, to assist other saliency cues such as contrast, achieving state-of-the-art results. However, their usage of boundary prior is very simple, fragile, and the integration with other cues is mostly heuristic. In this work, we present new methods to address these issues. First, we propose a robust background measure, called boundary connectivity. It characterizes the spatial layout of image regions with respect to image boundaries and is much more robust. It has an intuitive geometrical interpretation and presents unique benefits that are absent in previous saliency measures. Second, we propose a principled optimization framework to integrate multiple low level cues, including our background measure, to obtain clean and uniform saliency maps. Our formulation is intuitive, efficient and achieves state-of-the-art results on several benchmark datasets.

Product quantization (PQ) is an effective vector quantization method. A product quantizer can generate an exponentially large codebook at very low memory/time cost. The essence of PQ is to decompose the high-dimensional vector space into the Cartesian product of subspaces and then quantize these subspaces separately. The optimal space decomposition is important for the PQ performance, but still remains an unaddressed issue. In this paper, we optimize PQ by minimizing quantization distortions w.r.t the space decomposition and the quantization codebooks. We present two novel solutions to this challenging optimization problem. The first solution iteratively solves two simpler sub-problems. The second solution is based on a Gaussian assumption and provides theoretical analysis of the optimality. We evaluate our optimized product quantizers in three applications: (i) compact encoding for exhaustive ranking [1], (ii) building inverted multi-indexing for non-exhaustive search [2], and (iii) compacting image representations for image retrieval [3]. In all applications our optimized product quantizers outperform existing solutions.

Image completion involves filling missing parts in images. In this paper we address this problem through novel statistics of similar patches. We observe that if we match similar patches in the image and obtain their offsets (relative positions), the statistics of these offsets are sparsely distributed. We further observe that a few dominant offsets provide reliable information for completing the image. Such statistics can be incorporated into both matching-based and graph-based methods for image completion. Experiments show that our method yields better results in various challenging cases, and is faster than existing state-of-the-art methods.

This paper presents a fast denoising method that produces a clean image from a burst of noisy images. We accelerate alignment of the images by introducing a lightweight camera motion representation called homography flow . The aligned images are then fused to create a denoised output with rapid per-pixel operations in temporal and spatial domains. To handle scene motion during the capture, a mechanism of selecting consistent pixels for temporal fusion is proposed to "synthesize" a clean, ghost-free image, which can largely reduce the computation of tracking motion between frames. Combined with these efficient solutions, our method runs several orders of magnitude faster than previous work, while the denoising quality is comparable. A smartphone prototype demonstrates that our method is practical and works well on a large variety of real examples.

Systematic summarizing clinical efficacies of wrinkled giant hyssop to provide references for clinician's medication. Clinical efficacies of wrinkled giant hyssop can be summarized as: dissolving damp with aromatics, regulating qi to soothe zhongjiao (the middle heater); relieve summer-heat and cure cholera, expel dampness and relieving the exterior; militating zhongjiao (the middle heater) to stop vomitting, warming spleen to relieve pain; regulating qi to prevent miscarriage, relieving asthma by depressing upward-reverse flow of qi; oral deodorization, relieve swelling and hastening rupture.

The bag-of-features(BOF) image representation [7] is popular in largescale image retrieval. With BOF, the memory to store the inverted index file and the search complexity are both approximately linearly increased with the number of images. To address the retrieval efficiency and the memory constraint problem, besides some improvement work based on BOF, there come alternative approaches which aggregate local descriptors in one image into a single vector using Fisher Vector [6] or Vector of Local Aggregated Descriptor (VLAD) [1]. It has been shown in [1] that with as few as 16 bytes to represent an image, the retrieval performance is still comparable to that of the BOF representation. In this paper, we illustrate that Fisher Vector, VLAD and BOF can be uniformly derived in two steps: i Encoding – separately map each local descriptor into a code, and ii Pooling – aggregate all codes from one image into a single vector. Motivated by the success of these two-step approaches, we propose to use sparse coding(SC) framework to aggregate local feature for image retrieval. SC framework is firstly introduced by [10] for the task of image classification. It is a classical two-step approach: Step 1: Encoding. Each local descriptor x from an image is encoded into an N-dimensional vector u = [u1,u2, ...,uN ] by fitting a linear model with sparsity (L1) constraint:

Making a high-dimensional (e.g., 100K-dim) feature for face recognition seems not a good idea because it will bring difficulties on consequent training, computation, and storage. This prevents further exploration of the use of a high dimensional feature. In this paper, we study the performance of a high dimensional feature. We first empirically show that high dimensionality is critical to high performance. A 100K-dim feature, based on a single-type Local Binary Pattern (LBP) descriptor, can achieve significant improvements over both its low-dimensional version and the state-of-the-art. We also make the high-dimensional feature practical. With our proposed sparse projection method, named rotated sparse regression, both computation and model storage can be reduced by over 100 times without sacrificing accuracy quality.

Product quantization is an effective vector quantization approach to compactly encode high-dimensional vectors for fast approximate nearest neighbor (ANN) search. The essence of product quantization is to decompose the original high-dimensional space into the Cartesian product of a finite number of low-dimensional subspaces that are then quantized separately. Optimal space decomposition is important for the performance of ANN search, but still remains unaddressed. In this paper, we optimize product quantization by minimizing quantization distortions w.r.t. the space decomposition and the quantization codebooks. We present two novel methods for optimization: a non-parametric method that alternatively solves two smaller sub-problems, and a parametric method that is guaranteed to achieve the optimal solution if the input data follows some Gaussian distribution. We show by experiments that our optimized approach substantially improves the accuracy of product quantization for ANN search.

In computer vision there has been increasing interest in learning hashing codes whose Hamming distance approximates the data similarity. The hashing functions play roles in both quantizing the vector space and generating similarity-preserving codes. Most existing hashing methods use hyper-planes (or kernelized hyper-planes) to quantize and encode. In this paper, we present a hashing method adopting the k-means quantization. We propose a novel Affinity-Preserving K-means algorithm which simultaneously performs k-means clustering and learns the binary indices of the quantized cells. The distance between the cells is approximated by the Hamming distance of the cell indices. We further generalize our algorithm to a product space for learning longer codes. Experiments show our method, named as K-means Hashing (KMH), outperforms various state-of-the-art hashing encoding methods.

Face verification involves determining whether a pair of facial images belongs to the same or different subjects. This problem can prove to be quite challenging in many important applications where labeled training data is scarce, e.g., family album photo organization software. Herein we propose a principled transfer learning approach for merging plentiful source-domain data with limited samples from some target domain of interest to create a classifier that ideally performs nearly as well as if rich target-domain data were present. Based upon a surprisingly simple generative Bayesian model, our approach combines a KL-divergence based regularizer/prior with a robust likelihood function leading to a scalable implementation via the EM algorithm. As justification for our design choices, we later use principles from convex analysis to recast our algorithm as an equivalent structured rank minimization problem leading to a number of interesting insights related to solution structure and feature-transform invariance. These insights help to both explain the effectiveness of our algorithm as well as elucidate a wide variety of related Bayesian approaches. Experimental testing with challenging datasets validate the utility of the proposed algorithm.

We present an image editing tool called Content-Aware Rotation. Casually shot photos can appear tilted, and are often corrected by rotation and cropping. This trivial solution may remove desired content and hurt image integrity. Instead of doing rigid rotation, we propose a warping method that creates the perception of rotation and avoids cropping. Human vision studies suggest that the perception of rotation is mainly due to horizontal/vertical lines. We design an optimization-based method that preserves the rotation of horizontal/vertical lines, maintains the completeness of the image content, and reduces the warping distortion. An efficient algorithm is developed to address the challenging optimization. We demonstrate our content-aware rotation method on a variety of practical cases.

Despite the continuous advances in local stereo matching for years, most efforts are on developing robust cost computation and aggregation methods. Little attention has been seriously paid to the disparity refinement. In this work, we study weighted median filtering for disparity refinement. We discover that with this refinement, even the simple box filter aggregation achieves comparable accuracy with various sophisticated aggregation methods (with the same refinement). This is due to the nice weighted median filtering properties of removing outlier error while respecting edges/structures. This reveals that the previously overlooked refinement can be at least as crucial as aggregation. We also develop the first constant time algorithm for the previously time-consuming weighted median filter. This makes the simple combination ``box aggregation + weighted median'' an attractive solution in practice for both speed and accuracy. As a byproduct, the fast weighted median filtering unleashes its potential in other applications that were hampered by high complexities. We show its superiority in various applications such as depth up sampling, clip-art JPEG artifact removal, and image stylization.

Inverted indexing is a popular non-exhaustive solution to large scale search. An inverted file is built by a quantizer such as k-means or a tree structure. It has been found that multiple inverted files, obtained by multiple independent random quantizers, are able to achieve practically good recall and speed. Instead of computing the multiple quantizers independently, we present a method that creates them jointly. Our method jointly optimizes all code words in all quantizers. Then it assigns these code words to the quantizers. In experiments this method shows significant improvement over various existing methods that use multiple independent quantizers. On the one-billion set of SIFT vectors, our method is faster and more accurate than a recent state-of-the-art inverted indexing method.

In this paper, we propose a novel explicit image filter called guided filter. Derived from a local linear model, the guided filter computes the filtering output by considering the content of a guidance image, which can be the input image itself or another different image. The guided filter can be used as an edge-preserving smoothing operator like the popular bilateral filter [1], but it has better behaviors near edges. The guided filter is also a more generic concept beyond smoothing: It can transfer the structures of the guidance image to the filtering output, enabling new filtering applications like dehazing and guided feathering. Moreover, the guided filter naturally has a fast and nonapproximate linear time algorithm, regardless of the kernel size and the intensity range. Currently, it is one of the fastest edge-preserving filters. Experiments show that the guided filter is both effective and efficient in a great variety of computer vision and computer graphics applications, including edge-aware smoothing, detail enhancement, HDR compression, image matting/feathering, dehazing, joint upsampling, etc.

Stitched panoramic images mostly have irregular boundaries. Artists and common users generally prefer rectangular boundaries, which can be obtained through cropping or image completion techniques. In this paper, we present a content-aware warping algorithm that generates rectangular images from stitched panoramic images. Our algorithm consists of two steps. The first local step is mesh-free and preliminarily warps the image into a rectangle. With a grid mesh placed on this rectangle, the second global step optimizes the mesh to preserve shapes and straight lines. In various experiments we demonstrate that the results of our approach are often visually plausible, and the introduced distortion is often unnoticeable.

We present a novel video stabilization method which models camera motion with a bundle of (multiple) camera paths. The proposed model is based on a mesh-based, spatially-variant motion representation and an adaptive, space-time path optimization. Our motion representation allows us to fundamentally handle parallax and rolling shutter effects while it does not require long feature trajectories or sparse 3D reconstruction. We introduce the 'as-similar-as-possible' idea to make motion estimation more robust. Our space-time path smoothing adaptively adjusts smoothness strength by considering discontinuities, cropping size and geometrical distortion in a unified optimization framework. The evaluation on a large variety of consumer videos demonstrates the merits of our method.

This paper evaluates the usefulness of different simulators developed in server system testing for a mobile application during 5 years, with the purpose of identifying the advantages and weaknesses of each simulation method in the given setup. The study found that simulation was used on almost every interface to external systems, and two factors that affected the usefulness of simulators over long term were the functionality supported and the applicable usage scenarios. Large-scale simulation did not keep yielding high returns. In the mobile application of server/client architecture, emulating external servers raised the testing capability, and emulating mobile clients led to higher efficiency. The findings in this study can be used as references for software teams to plan simulation in QA activities properly for faster QA cycles, deeper understanding of the system, more stable test results and lower resource costs.

We present a very efficient, highly accurate, “Explicit Shape Regression” approach for face alignment. Unlike previous regression-based approaches, we directly learn a vectorial regression function to infer the whole facial shape (a set of facial landmarks) from the image and explicitly minimize the alignment errors over the training data. The inherent shape constraint is naturally encoded into the regressor in a cascaded learning framework and applied from coarse to fine during the test, without using a fixed parametric shape model as in most previous methods. To make the regression more effective and efficient, we design a two-level boosted regression, shape-indexed features and a correlation-based feature selection method. This combination enables us to learn accurate models from large training data in a short time (20 minutes for 2,000 training images), and run regression extremely fast in test (15 ms for a 87 landmarks shape). Experiments on challenging data show that our approach significantly outperforms the state-of-the-art in terms of both accuracy and efficiency.

Matching patches between two images, also known as computing nearest-neighbor fields, has been proven a useful technique in various computer vision/graphics algorithms. But this is a computationally challenging nearest-neighbor search task, because both the query set and the candidate set are of image size. In this paper, we propose Propagation-Assisted KD-Trees to quickly compute an approximate solution. We develop a novel propagation search method for kd-trees. In this method the tree nodes checked by each query are propagated from the nearby queries. This method not only avoids the time-consuming backtracking in traditional tree methods, but is more accurate. Experiments on public data show that our method is 10-20 times faster than the PatchMatch method [4] at the same accuracy, or reduces its error by 70% at the same running time. Our method is also 2-5 times faster and is more accurate than Coherency Sensitive Hashing [22], a latest state-of-the-art method.

Previous video stabilization methods often employ homographies to model transitions between consecutive frames, or require robust long feature tracks. However, the homography model is invalid for scenes with significant depth variations, and feature point tracking is fragile in videos with textureless objects, severe occlusion or camera rotation. To address these challenging cases, we propose to solve video stabilization with an additional depth sensor such as the Kinect camera. Though the depth image is noisy, incomplete and low resolution, it facilitates both camera motion estimation and frame warping, which make the video stabilization a much well posed problem. The experiments demonstrate the effectiveness of our algorithm.

For scene classification, patch-level linear features do not always work as well as handcrafted features. In this paper, we present a new model to greatly improve the usefulness of linear features in classification by introducing co-variance patterns. We analyze their properties, discuss the fundamental importance, and present a generative model to properly utilize them. With this set of covariance information, in our framework, even the most naive linear features that originally lack the vital ability in classification become powerful. Experiments show that the performance of our new covariance model based on linear features is comparable with or even better than handcrafted features in scene classification.

Alpha matting refers to the problem of softly extracting the foreground from an image. Given a trimap (specifying known foreground/background and unknown pixels), a straightforward way to compute the alpha value is to sample some known foreground and background colors for each unknown pixel. Existing sampling-based matting methods often collect samples near the unknown pixels only. They fail if good samples cannot be found nearby. In this paper, we propose a global sampling method that uses all samples available in the image. Our global sample set avoids missing good samples. A simple but effective cost function is defined to tackle the ambiguity in the sample selection process. To handle the computational complexity introduced by the large number of samples, we pose the sampling task as a correspondence problem. The correspondence search is efficiently achieved by generalizing a randomized algorithm previously designed for patch matching[3]. A variety of experiments show that our global sampling method produces both visually and quantitatively high-quality matting results.

Handling intra-personal variation is a major challenge in face recognition. It is difficult how to appropriately measure the similarity between human faces under significantly different settings (e.g., pose, illumination, and expression). In this paper, we propose a new model, called “Associate-Predict” (AP) model, to address this issue. The associate-predict model is built on an extra generic identity data set, in which each identity contains multiple images with large intra-personal variation. When considering two faces under significantly different settings (e.g., non-frontal and frontal), we first “associate” one input face with alike identities from the generic identity date set. Using the associated faces, we generatively “predict” the appearance of one input face under the setting of another input face, or discriminatively “predict” the likelihood whether two input faces are from the same person or not. We call the two proposed prediction methods as “appearance-prediction” and “likelihood-prediction”. By leveraging an extra data set (“memory”) and the “associate-predict” model, the intra-personal variation can be effectively handled. To improve the generalization ability of our model, we further add a switching mechanism - we directly compare the appearances of two faces if they have close intra-personal settings; otherwise, we use the associate-predict model for the recognition. Experiments on two public face benchmarks (Multi-PIE and LFW) demonstrated that our final model can substantially improve the performance of most existing face recognition methods.

We present a novel clustering algorithm for tagging a face dataset (e. g., a personal photo album). The core of the algorithm is a new dissimilarity, called Rank-Order distance, which measures the dissimilarity between two faces using their neighboring information in the dataset. The Rank-Order distance is motivated by an observation that faces of the same person usually share their top neighbors. Specifically, for each face, we generate a ranking order list by sorting all other faces in the dataset by absolute distance (e. g., L1 or L2 distance between extracted face recognition features). Then, the Rank-Order distance of two faces is calculated using their ranking orders. Using the new distance, a Rank-Order distance based clustering algorithm is designed to iteratively group all faces into a small number of clusters for effective tagging. The proposed algorithm outperforms competitive clustering algorithms in term of both precision/recall and efficiency.

Conventional saliency analysis methods measure the saliency of individual pixels. The resulting saliency map inevitably loses information in the original image and finding salient objects in it is difficult. We propose to detect salient objects by directly measuring the saliency of an image window in the original image and adopt the well established sliding window based object detection paradigm. We present a simple definition for window saliency, i.e., the cost of composing the window using the remaining parts of the image. The definition uses the entire image as the context and agrees with human intuition. It no longer relies on idealistic assumptions usually used before (e.g., "back- ground is homogenous") and generalizes well to complex objects and backgrounds in real world images. To realize the definition, we illustrate how to incorporate different cues such as appearance, position, and size. Based on a segment-based representation, the window composition cost function can be efficiently evaluated by a greedy optimization algorithm. Extensive evaluation on challenging object detection datasets verifies better efficacy and efficiency of the proposed method comparing to the state-of-the-art, making it a good pre-processing tool for subsequent applications. Moreover, we hope to stimulate further work towards the challenging yet important problem of generic salient object detection.

A camera RAW file contains minimally processed data from the image sensor. The contents of the RAW file include more information, and potentially higher quality, than the commonly used JPEG file. But the RAW file is typically several times larger than the JPEG file (taking fewer images, slower quick shooting) and lacks the standard file format (not ready-to-use, prolonging the image workflow). These drawbacks limit its applications. In this paper, we suggest a new “hybrid” image capture mode: a high-res JPEG file and a low-res RAW file as alternative of the original RAW file. Most RAW users can be benefited from such a combination. To address this problem, we provide an effective approach to reconstruct a high quality image by combining the advantages of two kinds of files. We formulate this reconstruction process as a global optimization problem by enforcing two constraints: reconstruction constraint and detail consistency constraint. The final recovered image is smaller than the full-res RAW file, enables faster quick shooting, and has both richer information (e.g., color space, dynamic range, lossless 14 bits data) and higher resolution. In practice, the functionality of capturing such a “hybrid” image pair in one-shot has been supported in some existing digital cameras.

In this paper, we propose a simple but effective image prior-dark channel prior to remove haze from a single input image. The dark channel prior is a kind of statistics of outdoor haze-free images. It is based on a key observation-most local patches in outdoor haze-free images contain some pixels whose intensity is very low in at least one color channel. Using this prior with the haze imaging model, we can directly estimate the thickness of the haze and recover a high-quality haze-free image. Results on a variety of hazy images demonstrate the power of the proposed prior. Moreover, a high-quality depth map can also be obtained as a byproduct of haze removal.

In this paper, we study the salient object detection problem for images. We formulate this problem as a binary labeling task where we separate the salient object from the background. We propose a set of novel features, including multiscale contrast, center-surround histogram, and color spatial distribution, to describe a salient object locally, regionally, and globally. A conditional random field is learned to effectively combine these features for salient object detection. Further, we extend the proposed approach to detect a salient object from sequential images by introducing the dynamic salient features. We collected a large image database containing tens of thousands of carefully labeled images by multiple users and a video segment database, and conducted a set of experiments over them to demonstrate the effectiveness of the proposed approach.

State-of-the-art image retrieval systems achieve scalability by using a bag-of-words representation and textual retrieval methods, but their performance degrades quickly in the face image domain, mainly because they produce visual words with low discriminative power for face images and ignore the special properties of faces. The leading features for face recognition can achieve good retrieval performance, but these features are not suitable for inverted indexing as they are high-dimensional and global and thus not scalable in either computational or storage cost. In this paper, we aim to build a scalable face image retrieval system. For this purpose, we develop a new scalable face representation using both local and global features. In the indexing stage, we exploit special properties of faces to design new component-based local features, which are subsequently quantized into visual words using a novel identity-based quantization scheme. We also use a very small Hamming signature (40 bytes) to encode the discriminative global feature for each face. In the retrieval stage, candidate images are first retrieved from the inverted index of visual words. We then use a new multireference distance to rerank the candidate images using the Hamming signature. On a one millon face database, we show that our local features and global Hamming signatures are complementary--the inverted index based on local features provides candidate images with good recall, while the multireference reranking with global Hamming signature leads to good precision. As a result, our system is not only scalable but also outperforms the linear scan retrieval system using the state-of-the-art face recognition feature in term of the quality.

In this paper, we propose a novel generic image prior-gradient profile prior, which implies the prior knowledge of natural image gradients. In this prior, the image gradients are represented by gradient profiles, which are 1-D profiles of gradient magnitudes perpendicular to image structures. We model the gradient profiles by a parametric gradient profile model. Using this model, the prior knowledge of the gradient profiles are learned from a large collection of natural images, which are called gradient profile prior. Based on this prior, we propose a gradient field transformation to constrain the gradient fields of the high resolution image and the enhanced image when performing single image super-resolution and sharpness enhancement. With this simple but very effective approach, we are able to produce state-of-the-art results. The reconstructed high resolution images or the enhanced images are sharp while have rare ringing or jaggy artifacts.

Robust object detection has many important applications in real-world online photo processing. For example, both Google image search and MSN live image search have integrated human face detector to retrieve face or portrait photos. Inspired by the success of such face filtering approach, in this paper, we focus on another popular online photo category--animal, which is one of the top five categories in the MSN live image search query log. As a first attempt, we focus on the problem of animal head detection of a set of relatively large land animals that are popular on the internet, such as cat, tiger, panda, fox, and cheetah. First, we proposed a new set of gradient oriented feature, Haar of Oriented Gradients (HOOG), to effectively capture the shape and texture features on animal head. Then, we proposed two detection algorithms, namely Bruteforce detection and Deformable detection, to effectively exploit the shape feature and texture feature simultaneously. Experimental results on 14,379 well labeled animals images validate the superiority of the proposed approach. Additionally, we apply the animal head detector to improve the image search result through text based online photo search result filtering.

We present a novel approach to address the representation issue and the matching issue in face recognition (verification). Firstly, our approach encodes the micro-structures of the face by a new learning-based encoding method. Unlike many previous manually designed encoding methods (e.g., LBP or SIFT), we use unsupervised learning techniques to learn an encoder from the training examples, which can automatically achieve very good tradeoff between discriminative power and invariance. Then we apply PCA to get a compact face descriptor. We find that a simple normalization mechanism after PCA can further improve the discriminative ability of the descriptor. The resulting face representation, learning-based (LE) descriptor, is compact, highly discriminative, and easy-to-extract. To handle the large pose variation in real-life scenarios, we propose a pose-adaptive matching method that uses pose-specific classifiers to deal with different pose combinations (e.g., frontal v.s. frontal, frontal v.s. left) of the matching face pair. Our approach is comparable with the state-of-the-art methods on the Labeled Face in Wild (LFW) benchmark (we achieved 84.45% recognition rate), while maintaining excellent compactness, simplicity, and generalization aability across different datasets.

Image matting is of great importance in both computer vision and graphics applications. Most existing state-of-the-art techniques rely on large sparse matrices such as the matting Laplacian. However, solving these linear systems is often time-consuming, which is unfavored for the user interaction. In this paper, we propose a fast method for high quality matting. We first derive an efficient algorithm to solve a large kernel matting Laplacian. A large kernel propagates information more quickly and may improve the matte quality. To further reduce running time, we also use adaptive kernel sizes by a KD-tree trimap segmentation technique. A variety of experiments show that our algorithm provides high quality results and is 5 to 20 times faster than previous methods.

Graph-cuts optimization is prevalent in vision and graphics problems. It is thus of great practical importance to parallelize the graph-cuts optimization using today's ubiquitous multi-core machines. However, the current best serial algorithm by Boykov and Kolmogorov (called the BK algorithm) still has the superior empirical performance. It is non-trivial to parallelize as expensive synchronization overhead easily offsets the advantage of parallelism. In this paper, we propose a novel adaptive bottom-up approach to parallelize the BK algorithm. We first uniformly partition the graph into a number of regularly-shaped disjoint subgraphs and process them in parallel, then we incrementally merge the subgraphs in an adaptive way to obtain the global optimum. The new algorithm has three benefits: 1) it is more cache-friendly within smaller subgraphs; 2) it keeps balanced workloads among computing cores; 3) it causes little overhead and is adaptable to the number of available cores. Extensive experiments in common applications such as 2D/3D image segmentations and 3D surface fitting demonstrate the effectiveness of our approach.

State-of-the-art image retrieval systems achieve scalability by using bag-of-words representation and textual retrieval methods, but their performance degrades quickly in the face image domain, mainly because they 1) produce visual words with low discriminative power for face images, and 2) ignore the special properties of the faces. The leading features for face recognition can achieve good retrieval performance, but these features are not suitable for inverted indexing as they are high-dimensional and global, thus not scalable in either computational or storage cost. In this paper we aim to build a scalable face image retrieval system. For this purpose, we develop a new scalable face representation using both local and global features. In the indexing stage, we exploit special properties of faces to design new component-based local features, which are subsequently quantized into visual words using a novel identity-based quantization scheme. We also use a very small hamming signature (40 bytes) to encode the discriminative global feature for each face. In the retrieval stage, candidate images are firstly retrieved from the inverted index of visual words. We then use a new multi-reference distance to re-rank the candidate images using the hamming signature. On a one-millon face database, we show that our local features and global hamming signatures are complementary - the inverted index based on local features provides candidate images with good recall, while the multi-reference re-ranking with global hamming signature leads to good precision. As a result, our system is not only scalable but also outperforms the linear scan retrieval system using the state-of-the-art face recognition feature in term of the quality.

Active Appearance Model (AAM) based face tracking has advantages of accurate alignment, high efficiency, and effectiveness for handling face deformation. However, AAM suffers from the generalization problem and has difficulties in images with cluttered backgrounds. In this paper, we introduce two novel constraints into AAM fitting to address the above problems. We first introduce a temporal matching constraint in AAM fitting. In the proposed fitting scheme, the temporal matching enforces an inter-frame local appearance constraint between frames. The resulting model takes advantage of temporal matching's good generalizability, but does not suffer from the mismatched points. To make AAM more stable for cluttered backgrounds, we introduce a color-based face segmentation as a soft constraint. Both constraints effectively improve the AAM tracker's performance, as demonstrated with experiments on various challenging real-world videos.

In this paper, we propose a simple but effective image prior - dark channel prior to remove haze from a single input image. The dark channel prior is a kind of statistics of the haze-free outdoor images. It is based on a key observation - most local patches in haze-free outdoor images contain some pixels which have very low intensities in at least one color channel. Using this prior with the haze imaging model, we can directly estimate the thickness of the haze and recover a high quality haze-free image. Results on a variety of outdoor haze images demonstrate the power of the proposed prior. Moreover, a high quality depth map can also be obtained as a by-product of haze removal.

In state-of-the-art image retrieval systems, an image is represented by a bag of visual words obtained by quantizing high-dimensional local image descriptors, and scalable schemes inspired by text retrieval are then applied for large scale image indexing and retrieval. Bag-of-words representations, however: 1) reduce the discriminative power of image features due to feature quantization; and 2) ignore geometric relationships among visual words. Exploiting such geometric constraints, by estimating a 2D affine transformation between a query image and each candidate image, has been shown to greatly improve retrieval precision but at high computational cost. In this paper we present a novel scheme where image features are bundled into local groups. Each group of bundled features becomes much more discriminative than a single feature, and within each group simple and robust geometric constraints can be efficiently enforced. Experiments in Web image search, with a database of more than one million images, show that our scheme achieves a 49% improvement in average precision over the baseline bag-of-words approach. Retrieval performance is comparable to existing full geometric verification approaches while being much less computationally expensive. When combined with full geometric verification we achieve a 77% precision improvement over the baseline bag-of-words approach, and a 24% improvement over full geometric verification alone.

The state-of-the-art content based image retrieval systems has been significantly advanced by the introduction of SIFT features and the bag-of-words image representation. Converting an image into a bag-of-words, however, involves three non-trivial steps: feature detection, feature description, and feature quantization. At each of these steps, there is a significant amount of information lost, and the resulted visual words are often not discriminative enough for large scale image retrieval applications. In this paper, we propose a novel multi-sample multi-tree approach to computing the visual word codebook. By encoding more information of the original image feature, our approach generates a much more discriminative visual word codebook that is also efficient in terms of both computation and space consumption, without losing the original repeatability of the visual features. We evaluate our approach using both a ground-truth data set and a real-world large scale image database. Our results show that a significant improvement in both precision and recall can be achieved by using the codebook derived from our approach.

In this paper, we address a novel problem of automatically creating a picture collage from a group of images. Picture collage is a kind of visual image summary-to arrange all input images on a given canvas, allowing overlay, to maximize visible visual information. We formulate the picture collage creation problem in a conditional random field model, which integrates image salience, canvas constraint, natural preference, and user interaction. Each image is represented by a group of weighted rectangles, which indicate the salient regions. Then picture collage is resolved by minimizing the energy, guided by the constraints. A two-step optimization method is proposed. First, a quick initialization algorithm based on the proposed 1D collage method is presented. Second, a very efficient Markov chain Monte Carlo method is designed for the refined optimization. We also integrate user interaction in the formulation and optimization to obtain an interactive collage reflecting personalized preference. Visual and quantitative experimental evaluations indicate the efficiency of the proposed collage creation technique.

In this paper, we present Paint Selection, a progressive painting-based tool for local selection in images. Paint Selection facilitates users to progressively make a selection by roughly painting the object of interest using a brush. More importantly, Paint Selection is efficient enough that instant feedback can be provided to users as they drag the mouse. We demonstrate that high quality selections can be quickly and effectively "painted" on a variety of multi-megapixel images.

In this paper, we present SkyFinder, an interactive search system of over a half million sky images downloaded from the Internet. Using a set of automatically extracted, semanticsky attributes(category, layout, richness, horizon, etc.), the user can find a desired sky image, such as "a landscape with rich clouds at sunset" or "a whole blue sky with white clouds". The system is fully automatic and scalable. It computes all sky attributes offline, then provides an interactive online search engine. Moreover, we build a sky graph based on the sky attributes, so that the user can smoothly explore and find a path within the space of skies. We also show how our system can be used for controllable sky replacement.

We present an interactive system for reconstructing surface normals from a single image. Our approach has two complementary contributions. First, we introduce a novel shape-from-shading algorithm (SfS) that produces faithful normal reconstruction for local image region (high-frequency component), but it fails to faithfully recover the overall global structure (low-frequency component). Our second contribution consists of an approach that corrects low-frequency error using a simple markup procedure. This approach, aptly calledrotation palette, allows the user to specify large scale corrections of surface normals by drawing simple stroke correspondences between the normal map and a sphere image which represents rotation directions. Combining these two approaches, we can produce high-quality surfaces quickly from single images.

Ringing is the most disturbing artifact in the image deconvolution. In this paper, we present a progressive inter-scale and intra-scale non-blind image deconvolution approach that significantly reduces ringing. Our approach is built on a novel edge-preserving deconvolution algorithm called bilateral Richardson-Lucy (BRL) which uses a large spatial support to handle large blur. We progressively recover the image from a coarse scale to a fine scale (inter-scale), and progressively restore image details within every scale (intra-scale). To perform the inter-scale deconvolution, we propose a joint bilateral Richardson-Lucy (JBRL) algorithm so that the recovered image in one scale can guide the deconvolution in the next scale. In each scale, we propose an iterative residual deconvolution to progressively recover image details. The experimental results show that our progressive deconvolution can produce images with very little ringing for large blur kernels.

We study visual attention by detecting a salient object in an input image. We formulate salient object detection as an image segmentation problem, where we separate the salient object from the image background. We propose a set of novel features including multi-scale contrast, center-surround histogram, and color spatial distribution to describe a salient object locally, regionally, and globally. A conditional random field is learned to effectively combine these features for salient object detection. We also constructed a large image database containing tens of thousands of carefully labeled images by multiple users. To our knowledge, it is the first large image database for quantitative evaluation of visual attention algorithms. We validate our approach on this image database, which is public available with this paper.

In this paper, we propose a novel approach for foreground layer extraction using flash/no-flash image pairs, which we call flash cut. Flash cut is based on the simple observation that only the foreground is significantly brightened by the flash and the background appearance change is very small, if the background is distant. Changes due to flash, motion, and color information are fused in an MRF framework to produce high quality segmentation results. Flash cut handles some amount of camera shake, and foreground motion, which makes it practical for anyone with a flash-equipped camera to use. We validate our approach on a variety of indoor and outdoor examples.

In this paper, we present an interactive offline tracking system for generic color objects. The system achieves 60- 100 fps on a 320 times 240 video. The user can therefore easily refine the tracking result in an interactive way. To fully exploit user input and reduce user interaction, the tracking problem is addressed in a global optimization framework. The optimization is efficiently performed through three steps. First, from user's input we train a fast object detector that locates candidate objects in the video based on proposed features called boosted color bin. Second, we exploit the temporal coherence to generate multiple object trajectories based on a global best-first strategy. Last, an optimal object path is found by dynamic programming.

Aligning a pair of blurred and non-blurred images is a prerequisite for many image and video restoration and graphics applications. The traditional alignment methods such as direct and feature-based approaches cannot be used due to the presence of motion blur in one image of the pair. In this paper, we present an effective and accurate alignment approach for a blurred/non-blurred image pair. We exploit a statistical characteristic of the real blur kernel - the marginal distribution of kernel value is sparse. Using this sparseness prior, we can search the best alignment which produces the sparsest blur kernel. The search is carried out in scale space with a coarse-to-fine strategy for efficiency. Finally, we demonstrate the effectiveness of our algorithm for image deblurring, video restoration, and image matting.

Recently, gradient meshes have been introduced as a powerful vector graphics representation to draw multicolored mesh objects with smooth transitions. Using tools from Abode Illustrator and Corel CorelDraw, a user can manually create gradient meshes even for photo-realistic vector arts, which can be further edited, stylized and animated. In this paper, we present an easy-to-use interactive tool, called optimized gradient mesh , to semi-automatically and quickly create gradient meshes from a raster image. We obtain the optimized gradient mesh by formulating an energy minimization problem. The user can also interactively specify a few vector lines to guide the mesh generation. The resulting optimized gradient mesh is an editable and scalable mesh that otherwise would have taken many hours for a user to manually create.

Taking satisfactory photos under dim lighting conditions using a hand-held camera is challenging. If the camera is set to a long exposure time, the image is blurred due to camera shake. On the other hand, the image is dark and noisy if it is taken with a short exposure time but with a high camera gain. By combining information extracted from both blurred and noisy images, however, we show in this paper how to produce a high quality image that cannot be obtained by simply denoising the noisy image, or deblurring the blurred image alone. Our approach is image deblurring with the help of the noisy image. First, both images are used to estimate an accurate blur kernel, which otherwise is difficult to obtain from a single blurred image. Second, and again using both images, a residual deconvolution is proposed to significantly reduce ringing artifacts inherent to image deconvolution. Third, the remaining ringing artifacts in smooth image regions are further suppressed by a gain-controlled deconvolution process. We demonstrate the effectiveness of our approach using a number of indoor and outdoor images taken by off-the-shelf hand-held cameras in poor lighting environments.

In this paper, we address a novel problem of automatically creating a picture collage from a group of images. Picture collage is a kind of visual image summary - to arrange all input images on a given canvas, allowing overlay, to maximize visible visual information. We formulate the picture collage creation problem in a Bayesian framework. The salient regions of each image are firstly extracted and represented as a set of weighted rectangles. Then, the image arrangement is formulated as a Maximum a Posterior (MAP) problem such that the output picture collage shows as many visible salient regions (without being overlaid by others) from all images as possible. Moreover, a very efficientMarkov chain Monte Carlo (MCMC) method is designed for the optimization. Applications to desktop image browsing and image search result summarization demonstrate the effectiveness of our approach.

In this paper, we present a user-friendly system for seamless image composition, which we call drag-and-drop pasting. We observe that for Poisson image editing [Perez et al. 2003] to work well, the user must carefully draw a boundary on the source image to indicate the region of interest, such that salient structures in source and target images do not conflict with each other along the boundary. To make Poisson image editing more practical and easy to use, we propose a new objective function to compute an optimized boundary condition. A shortest closed-path algorithm is designed to search for the location of the boundary. Moreover, to faithfully preserve the object's fractional boundary, we construct a blended guidance field to incorporate the object's alpha matte. To use our system, the user needs only to simply outline a region of interest in the source image, and then drag and drop it onto the target image. Experimental results demonstrate the effectiveness of our "drag-and-drop pasting" system.

In this paper, we propose a novel approach to extract mattes using a pair of flash/no-flash images. Our approach, which we call flash matting , was inspired by the simple observation that the most noticeable difference between the flash and no-flash images is the foreground object if the background scene is sufficiently distant. We apply a new matting algorithm called joint Bayesian flash matting to robustly recover the matte from flash/no-flash images, even for scenes in which the foreground and the background are similar or the background is complex. Experimental results involving a variety of complex indoors and outdoors scenes show that it is easy to extract high-quality mattes using an off-the-shelf, flash-equipped camera. We also describe extensions to flash matting for handling more general scenes.

In this paper, we propose a symmetric stereo model to handle occlusion in dense two-frame stereo. Our occlusion reasoning is directly based on the visibility constraint that is more general than both ordering and uniqueness constraints used in previous work. The visibility constraint requires occlusion in one image and disparity in the other to be consistent. We embed the visibility constraint within an energy minimization framework, resulting in a symmetric stereo model that treats left and right images equally. An iterative optimization algorithm is used to approximate the minimum of the energy using belief propagation. Our stereo model can also incorporate segmentation as a soft constraint. Experimental results on the Middlebury stereo images show that our algorithm is state-of-the-art.

In this paper, we present a novel approach to keyframe-based tracking, called bi-directional tracking. Given two object templates in the beginning and ending keyframes, the bi-directional tracker outputs the MAP (maximum a posterior) solution of the whole state sequence of the target object in the Bayesian framework. First, a number of 3D trajectory segments of the object are extracted from the input video, using a novel trajectory segment analysis. Second, these disconnected trajectory segments due to occlusion are linked by a number of inferred occlusion segments. Last, the MAP solution is obtained by trajectory optimization in a coarse-to-fine manner. Experimental results show the robustness of our approach with respect to sudden motion, ambiguity, and short and long periods of occlusion.

In this paper, we present a system for cutting a moving object out from a video clip. The cutout object sequence can be pasted onto another video or a background image. To achieve this, we first apply a new 3D graph cut based segmentation approach on the spatial-temporal video volume. Our algorithm partitions watershed presegmentation regions into foreground and background while preserving temporal coherence. Then, the initial segmentation result is refined locally. Given two frames in the video sequence, we specify two respective windows of interest which are then tracked using a bi-directional feature tracking algorithm. For each frame in between these two given frames, the segmentation in each tracked window is refined using a 2D graph cut that utilizes a local color model. Moreover, we provide brush tools for the user to control the object boundary precisely wherever needed. Based on the accurate binary segmentation result, we apply coherent matting to extract the alpha mattes and foreground colors of the object.

In this paper, we introduce a novel approach to image completion, which we call structure propagation. In our system, the user manually specifies important missing structure information by extending a few curves or line segments from the known to the unknown regions. Our approach synthesizes image patches along these user-specified curves in the unknown region using patches selected around the curves in the known region. Structure propagation is formulated as a global optimization problem by enforcing structure and consistency constraints. If only a single curve is specified, structure propagation is solved using Dynamic Programming. When multiple intersecting curves are specified, we adopt the Belief Propagation algorithm to find the optimal patches. After completing structure propagation, we fill in the remaining unknown regions using patch-based texture synthesis. We show that our approach works well on a number of examples that are challenging to state-of-the-art techniques.

In this paper, we present Lazy Snapping , an interactive image cutout tool. Lazy Snapping separates coarse and fine scale processing, making object specification and detailed adjustment easy . Moreover, Lazy Snapping provides instant visual feedback, snapping the cutout contour to the true object boundary efficiently despite the presence of ambiguous or low contrast edges. Instant feedback is made possible by a novel image segmentation algorithm which combines graph cut with pre-computed over-segmentation. A set of intuitive user interface (UI) tools is designed and implemented to provide flexible control and editing for the users. Usability studies indicate that Lazy Snapping provides a better user experience and produces better segmentation results than the state-of-the-art interactive image cutout tool, Magnetic Lasso in Adobe Photoshop.

In this article, we present an image-based modeling and rendering system, which we call pop-up light field , that models a sparse light field using a set of coherent layers . In our system, the user specifies how many coherent layers should be modeled or popped up according to the scene complexity. A coherent layer is defined as a collection of corresponding planar regions in the light field images. A coherent layer can be rendered free of aliasing all by itself, or against other background layers. To construct coherent layers, we introduce a Bayesian approach, coherence matting , to estimate alpha matting around segmented layer boundaries by incorporating a coherence prior in order to maintain coherence across images.We have developed an intuitive and easy-to-use user interface (UI) to facilitate pop-up light field construction. The key to our UI is the concept of human-in-the-loop where the user specifies where aliasing occurs in the rendered image. The user input is reflected in the input light field images where pop-up layers can be modified. The user feedback is instant through a hardware-accelerated real-time pop-up light field renderer. Experimental results demonstrate that our system is capable of rendering anti-aliased novel views from a sparse light field.

In this paper, we formulate the problem of natural image matting as one of solving Poisson equations with the matte gradient field. Our approach, which we call Poisson matting , has the following advantages. First, the matte is directly reconstructed from a continuous matte gradient field by solving Poisson equations using boundary information from a user-supplied trimap. Second, by interactively manipulating the matte gradient field using a number of filtering tools, the user can further improve Poisson matting results locally until he or she is satisfied. The modified local result is seamlessly integrated into the final result. Experiments on many complex natural images demonstrate that Poisson matting can generate good matting results that are not possible using existing matting techniques.

We propose a Bayesian approach to image hallucination. Given a generic low resolution image, we hallucinate a high resolution image using a set of training images. Our work is inspired by recent progress on natural image statistics that the priors of image primitives can be well represented by examples. Specifically, primal sketch priors (e.g., edges, ridges and corners) are constructed and used to enhance the quality of the hallucinated high resolution image. Moreover, a contour smoothness constraint enforces consistency of primitives in the hallucinated image by a Markov-chain based inference algorithm. A reconstruction constraint is also applied to further improve the quality of the hallucinated image. Experiments demonstrate that our approach can hallucinate high quality super-resolution images.

In this paper, we formulate the stereo matching problem as a Markov network and solve it using Bayesian belief propagation. The stereo Markov network consists of three coupled Markov random fields that model the following: a smooth field for depth/disparity, a line process for depth discontinuity, and a binary process for occlusion. After eliminating the line process and the binary process by introducing two robust functions, we apply the belief propagation algorithm to obtain the maximum a posteriori (MAP) estimation in the Markov network. Other low-level visual cues (e.g., image segmentation) can also be easily incorporated in our stereo model to obtain better stereo results. Experiments demonstrate that our methods are comparable to the state-of-the-art stereo algorithms for many test cases.