Diffusion models are highly effective at generating high-quality images but pose risks, such as the unintentional generation of NSFW (not safe for work) content. Although various techniques have been proposed to mitigate unwanted influences in diffusion models while preserving overall performance, achieving a balance between these goals remains challenging. In this work, we introduce EraseDiff, an algorithm designed to preserve the utility of the diffusion model on retained data while removing the unwanted information associated with the data to be forgotten. Our approach formulates this task as a constrained optimization problem using the value function, resulting in a natural first-order algorithm for solving the optimization problem. By altering the generative process to deviate away from the groundtruth denoising trajectory, we update parameters for preservation while controlling constraint reduction to ensure effective erasure, striking an optimal trade-off. Extensive experiments and thorough comparisons with state-of-the-art algorithms demonstrate that EraseDiff effectively preserves the model’s utility, efficacy, and efficiency.$\color{Red} {\text {WARNING}}$: This paper contains sexually explicit imagery that may be offensive in nature.

To efficiently adapt large models or to train generative models of neural representations, Hypernetworks have drawn interest. While hypernetworks work well, training them is cumbersome, and often requires ground truth optimized weights for each sample. However, obtaining each of these weights is a training problem of its own—one needs to train, e.g., adaptation weights or even an entire neural field for hypernetworks to regress to. In this work, we propose a method to train hypernetworks, without the need for any per-sample ground truth. Our key idea is to learn a Hypernetwork ‘Field’ and estimate the entire trajectory of network weight training instead of simply its converged state. In other words, we introduce an additional input to the Hypernetwork, the convergence state, which then makes it act as a neural field that models the entire convergence pathway of a task network. A critical benefit in doing so is that the gradient of the estimated weights at any convergence state must then match the gradients of the original task—this constraint alone is sufficient to train the Hypernetwork Field. We demonstrate the effectiveness of our method through the task of personalized image generation and 3D shape reconstruction from images and point clouds, demonstrating competitive results without any per-sample ground truth.

We propose a novel knowledge distillation approach, CustomKD, that effectively leverages large vision foundation models (LVFMs) to enhance the performance of edge models (e.g., MobileNetV3). Despite recent advancements in LVFMs, such as DINOv2 and CLIP, their potential in knowledge distillation for enhancing edge models remains underexplored. While knowledge distillation is a promising approach for improving the performance of edge models, the discrepancy in model capacities and heterogeneous architectures between LVFMs and edge models poses a significant challenge. Our observation indicates that although utilizing larger backbones (e.g., ViT-S to ViT-L) in teacher models improves their downstream task performances, the knowledge distillation from the large teacher models fails to bring as much performance gain for student models as for teacher models due to the large model discrepancy. Our simple yet effective CustomKD customizes the well-generalized features inherent in LVFMs to a given student model in order to reduce model discrepancies. Specifically, beyond providing well-generalized original knowledge from teachers, CustomKD aligns the features of teachers to those of students, making it easy for students to understand and overcome the large model discrepancy overall. CustomKD significantly improves the performances of edge models in scenarios with unlabeled data such as unsupervised domain adaptation (e.g., OfficeHome and DomainNet) and semi-supervised learning (e.g., CIFAR-100 and ImageNet), achieving the new state-of-the-art performances.

Few-shot class incremental learning (FSCIL) enables the continual learning of new concepts with only a few training examples. In FSCIL, the model undergoes substantial updates, making it prone to forgetting previous concepts and overfitting to the limited new examples. Most recent trend is typically to disentangle the learning of the representation from the classification head of the model. A well-generalized feature extractor on the base classes (many examples and many classes) is learned, and then fixed during incremental learning. Arguing that the fixed feature extractor restricts the model’s adaptability to new classes, we introduce a novel FSCIL method to effectively address catastrophic forgetting and overfitting issues. Our method enables to seamlessly update the entire model with a few examples. We mainly propose a tripartite weight-space ensemble (Tri-WE). Tri-WE interpolates the base, immediately previous, and current models in weight-space, especially for the classification heads of the models. Then, it collaboratively maintains knowledge from the base and previous models. In addition, we recognize the challenges of distilling generalized representations from the previous model from scarce data. Hence, we suggest a regularization loss term using amplified data knowledge distillation. Simply intermixing the few-shot data, we can produce richer data enabling the distillation of critical knowledge from the previous model. Consequently, we attain state-of-the-art results on the miniImageNet, CUB200, and CIFAR100 datasets.

We introduce Flashback Learning (FL), a novel method designed to harmonize the stability and plasticity of models in Continual Learning (CL). Unlike prior approaches that primarily focus on regularizing model updates to preserve old information while learning new concepts, FL explicitly balances this trade-off through a bidirectional form of regularization. This approach effectively guides the model to swiftly incorporate new knowledge while actively retaining its old knowledge. FL operates through a two-phase training process and can be seamlessly integrated into various CL methods, including replay, parameter regularization, distillation, and dynamic architecture techniques. In designing FL, we use two distinct knowledge bases: one to enhance plasticity and another to improve stability. FL ensures a more balanced model by utilizing both knowledge bases to regularize model updates. Theoretically, we analyze how the FL mechanism enhances the stability-plasticity balance. Empirically, FL demonstrates tangible improvements over baseline methods within the same training budget. By integrating FL into at least one representative baseline from each CL category, we observed an average accuracy improvement of up to 4.91% in Class-Incremental and 3.51% in Task-Incremental settings on standard image classification benchmarks. Additionally, measurements of the stability-to-plasticity ratio confirm that FL effectively enhances this balance. FL also outperforms state-of-the-art CL methods on more challenging datasets like ImageNet. The codes of this article will be available at https://github.com/csiro-robotics/Flashback-Learning.

Federated Learning (FL) is a distributed learning paradigm that enhances users' privacy by eliminating the need for clients to share raw, private data with the server. Despite the success, recent studies expose the vulnerability of FL to model inversion attacks, where adversaries reconstruct users’ private data via eavesdropping on the shared gradient information. We hypothesize that a key factor in the success of such attacks is the low entanglement among gradients per data within the batch during stochastic optimization. This creates a vulnerability that an adversary can exploit to reconstruct the sensitive data. Building upon this insight, we present a simple, yet effective defense strategy that obfuscates the gradients of the sensitive data with concealed samples. To achieve this, we propose synthesizing concealed samples to mimic the sensitive data at the gradient level while ensuring their visual dissimilarity from the actual sensitive data. Compared to the previous art, our empirical evaluations suggest that the proposed technique provides the strongest protection while simultaneously maintaining the FL performance. Code is located at https://github.com/JingWu321/DCS-2.

The scarcity of ground-truth labels poses one major challenge in developing optical flow estimation models that are both generalizable and robust. While current methods rely on data augmentation, they have yet to fully exploit the rich information available in labeled video sequences. We propose OCAI, a method that supports robust frame interpolation by generating intermediate video frames along-side optical flows in between. Utilizing a forward warping approach, OCAI employs occlusion awareness to resolve ambiguities in pixel values and fills in missing values by leveraging the forward-backward consistency of optical flows. Additionally, we introduce a teacher-student style semi-supervised learning method on top of the interpolated frames. Using a pair of unlabeled frames and the teacher model's predicted optical flow, we generate interpolated frames and flows to train a student model. The teacher's weights are maintained using Exponential Moving Averaging of the student. Our evaluations demonstrate perceptually superior interpolation quality and enhanced optical flow accuracy on established benchmarks such as Sintel and KITTI.

The detection and localization of highly realistic deepfake audio-visual content are challenging even for the most advanced state-of-the-art methods. While most of the research efforts in this domain are focused on detecting high-quality deepfake images and videos, only a few works address the problem of the localization of small segments of audio-visual manipulations embedded in real videos. In this research, we emulate the process of such content generation and propose the AV-Deepfake1M dataset. The dataset contains content-driven (i) video manipulations, (ii) audio manipulations, and (iii) audio-visual manipulations for more than 2K subjects resulting in a total of more than 1M videos. The paper provides a thorough description of the proposed data generation pipeline accompanied by a rigorous analysis of the quality of the generated data. The comprehensive benchmark of the proposed dataset utilizing state-of-the-art deepfake detection and localization methods indicates a significant drop in performance compared to previous datasets. The proposed dataset will play a vital role in building the next-generation deepfake localization methods. The dataset and associated code are available at https://github.com/ControlNet/AV-Deepfake1M.

The detection and localization of deepfake content, particularly when small fake segments are seamlessly mixed with real videos, remains a significant challenge in the field of digital media security. Based on the recently released AV-Deepfake1M dataset, which contains more than 1 million manipulated videos across more than 2,000 subjects, we introduce the 1M-Deepfakes Detection Challenge. This challenge is designed to engage the research community in developing advanced methods for detecting and localizing deepfake manipulations within the large-scale high-realistic audio-visual dataset. The participants can access the AV-Deepfake1M dataset and are required to submit their inference results for evaluation across the metrics for detection or localization tasks. The methodologies developed through the challenge will contribute to the development of next-generation deepfake detection and localization systems. Evaluation scripts, baseline models, and accompanying code will be available on https://github.com/ControlNet/AV-Deepfake1M.

While Low-Rank Adaptation (LoRA) has proven beneficial for efficiently fine-tuning large models, LoRA fine-tuned text-to-image diffusion models lack diversity in the generated images, as the model tends to copy data from the observed training samples. This effect becomes more pronounced at higher values of adapter strength and for adapters with higher ranks which are fine-tuned on smaller datasets. To address these challenges, we present FouRA, a novel low-rank method that learns projections in the Fourier domain along with learning a flexible input-dependent adapter rank selection strategy. Through extensive experiments and analysis, we show that FouRA successfully solves the problems related to data copying and distribution collapse while significantly improving the generated image quality. We demonstrate that FouRA enhances the generalization of fine-tuned models thanks to its adaptive rank selection. We further show that the learned projections in the frequency domain are decorrelated and prove effective when merging multiple adapters. While FouRA is motivated for vision tasks, we also demonstrate its merits for language tasks on commonsense reasoning and GLUE benchmarks.

Medical image synthesis represents a critical area of research in clinical decision-making, aiming to overcome the challenges associated with acquiring multiple image modalities for an accurate clinical workflow. This approach proves beneficial in estimating an image of a desired modality from a given source modality among the most common medical imaging contrasts, such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron Emission Tomography (PET). However, translating between two image modalities presents difficulties due to the complex and non-linear domain mappings. Deep learning-based generative modelling has exhibited superior performance in synthetic image contrast applications compared to conventional image synthesis methods. This survey comprehensively reviews deep learning-based medical imaging translation from 2018 to 2023 on pseudo-CT, synthetic MR, and synthetic PET. We provide an overview of synthetic contrasts in medical imaging and the most frequently employed deep learning networks for medical image synthesis. Additionally, we conduct a detailed analysis of each synthesis method, focusing on their diverse model designs based on input domains and network architectures. We also analyse novel network architectures, ranging from conventional CNNs to the recent Transformer and Diffusion models. This analysis includes comparing loss functions, available datasets and anatomical regions, and image quality assessments and performance in other downstream tasks. Finally, we discuss the challenges and identify solutions within the literature, suggesting possible future directions. We hope that the insights offered in this survey paper will serve as a valuable roadmap for researchers in the field of medical image synthesis.

Eye gaze analysis is an important research problem in the field of Computer Vision and Human-Computer Interaction. Even with notable progress in the last 10 years, automatic gaze analysis still remains challenging due to the uniqueness of eye appearance, eye-head interplay, occlusion, image quality, and illumination conditions. There are several open questions, including what are the important cues to interpret gaze direction in an unconstrained environment without prior knowledge and how to encode them in real-time. We review the progress across a range of gaze analysis tasks and applications to elucidate these fundamental questions, identify effective methods in gaze analysis, and provide possible future directions. We analyze recent gaze estimation and segmentation methods, especially in the unsupervised and weakly supervised domain, based on their advantages and reported evaluation metrics. Our analysis shows that the development of a robust and generic gaze analysis method still needs to address real-world challenges such as unconstrained setup and learning with less supervision. We conclude by discussing future research directions for designing a real-world gaze analysis system that can propagate to other domains including Computer Vision, Augmented Reality (AR), Virtual Reality (VR), and Human Computer Interaction (HCI).

In recent years, open-vocabulary (OV) dense visual prediction (such as OV object detection, semantic, instance and panoptic segmentations) has attracted increasing research attention. However, most of the existing approaches are task-specific, i.e., tackling each task individually. In this paper, we propose a Unified Open-Vocabulary Network (UOVN) to jointly address four common dense prediction tasks. Compared with separate models, a unified network is more desirable for diverse industrial applications. Moreover, OV dense prediction training data is relatively less. Separate networks can only leverage task-relevant training data, while a unified approach can integrate diverse data to boost individual tasks. We address two major challenges in unified OV prediction. Firstly, unlike unified methods for fixed-set predictions, OV networks are usually trained with multi-modal data. Therefore, we propose a multi-modal, multi-scale and multi-task (MMM) decoding mechanism to better exploit multi-modal information for OV recognition. Secondly, because UOVN uses data from different tasks for training, there are significant domain and task gaps. We present a UOVN training mechanism to reduce such gaps. Experiments on four datasets demonstrate the effectiveness of our UOVN.

This paper proposes a self-supervised approach to learn universal facial representations from videos, that can transfer across a variety of facial analysis tasks such as Facial Attribute Recognition (FAR), Facial Expression Recognition (FER), DeepFake Detection (DFD), and Lip Synchronization (LS). Our proposed framework, named MARLIN, is a facial video masked autoencoder, that learns highly robust and generic facial embeddings from abundantly available non-annotated web crawled facial videos. As a challenging auxiliary task, MARLIN reconstructs the spatio-temporal details of the face from the densely masked facial regions which mainly include eyes, nose, mouth, lips, and skin to capture local and global aspects that in turn help in encoding generic and transferable features. Through a variety of experiments on diverse downstream tasks, we demonstrate MARLIN to be an excellent facial video encoder as well as feature extractor, that performs consistently well across a variety of downstream tasks including FAR (1.13% gain over supervised benchmark), FER (2.64% gain over unsupervised benchmark), DFD (1.86% gain over unsupervised benchmark), LS (29.36% gain for Frechet Inception Distance), and even in low data regime. Our code and models are available at https://github.com/ControlNet/MARLIN.

3D visual language reasoning plays an important role in effective human-computer interaction. The current approaches for 3D visual reasoning are task-specific, and lack pre-training methods to learn generic representations that can transfer across various tasks. Despite the encouraging progress in vision-language pre-training for image-text data, 3D-language pre-training is still an open issue due to limited 3D-language paired data, highly sparse and irregular structure of point clouds and ambiguities in spatial relations of 3D objects with viewpoint changes. In this paper, we present a generic 3D-language pre-training approach, that tackles multiple facets of 3D-language reasoning by learning universal representations. Our learning objective constitutes two main parts. 1) Context aware spatial-semantic alignment to establish fine-grained correspondence between point clouds and texts. It reduces relational ambiguities by aligning 3D spatial relationships with textual semantic context. 2) Mutual 3D-Language Masked modeling to enable cross-modality information exchange. Instead of reconstructing sparse 3D points for which language can hardly provide cues, we propose masked proposal reasoning to learn semantic class and mask-invariant representations. Our proposed 3D-language pre-training method achieves promising results once adapted to various downstream tasks, including 3D visual grounding, 3D dense captioning and 3D question answering. Our codes are available at https://github.com/leolyj/3D-VLP

Confidence-based pseudo-labeling is among the dominant approaches in semi-supervised learning (SSL). It relies on including high-confidence predictions made on unlabeled data as additional targets to train the model. We propose Protocon, a novel SSL method aimed at the less-explored label-scarce SSL where such methods usually underperform. Protocon refines the pseudolabels by lever-aging their nearest neighbours' information. The neighbours are identified as the training proceeds using an online clustering approach operating in an embedding space trained via a prototypical loss to encourage well-formed clusters. The online nature of Protocon allows it to utilise the label history of the entire dataset in one training cycle to refine labels in the following cycle without the need to store image embeddings. Hence, it can seamlessly scale to larger datasets at a low cost. Finally, Protocon addresses the poor training signal in the initial phase of training (due to fewer confident predictions) by introducing an auxiliary self-supervised loss. It delivers significant gains and faster convergence over state-of-the-art across 5 datasets, including CIFARs, ImageNet and DomainNet.

Effectively encoding multi-scale contextual information is crucial for accurate semantic segmentation. Most of the existing transformer-based segmentation models combine features across scales without any selection, where features on sub-optimal scales may degrade segmentation outcomes. Leveraging from the inherent properties of Vision Transformers, we propose a simple yet effective module, Transformer Scale Gate (TSG), to optimally combine multi-scale features. TSG exploits cues in self and cross attentions in Vision Transformers for the scale selection. TSG is a highly flexible plug-and-play module, and can easily be incorporated with any encoder-decoder-based hierarchical vision Transformer. Extensive experiments on the Pascal Context, ADE20K and Cityscapes datasets demonstrate that the proposed feature selection strategy achieves consistent gains.

Despite the significant recent progress made on 3D point cloud semantic segmentation, the current methods require training data for all classes at once, and are not suitable for real-life scenarios where new categories are being continuously discovered. Substantial memory storage and expensive re-training is required to update the model to sequentially arriving data for new concepts. In this paper, to continually learn new categories using previous knowledge, we introduce class-incremental semantic segmentation of 3D point cloud. Unlike 2D images, 3D point clouds are disordered and unstructured, making it difficult to store and transfer knowledge especially when the previous data is not available. We further face the challenge of semantic shift, where previous/future classes are indiscriminately collapsed and treated as the background in the current step, causing a dramatic performance drop on past classes. We exploit the structure of point cloud and propose two strategies to address these challenges. First, we design a geometry-aware distillation module that transfers point-wise feature associations in terms of their geometric characteristics. To counter forgetting caused by the semantic shift, we further develop an uncertainty-aware pseudo-labelling scheme that eliminates noise in uncertain pseudo-labels by label propagation within a local neighborhood. Our extensive experiments on S3DIS and ScanNet in a class-incremental setting show impressive results comparable to the joint training strategy (upper bound). Code is available at: https://github.com/leolyj/3DPC-CISS

We propose an Unsupervised Domain Adaptation (UDA) method by making use of Energy-Based Learning (EBL) and demonstrate 1. EBL can be used to improve the instance selection for a self-training task on the unlabelled target domain, and 2. alignment and normalizing energy scores can learn domain-invariant representations. For the former, we show that an energy-based selection criterion can be used to model instance selections by mimicking the joint distribution between data and predictions in the target domain. As per learning domain invariant representations, we show that stable domain alignment can be achieved by a combined energy alignment and an energy normalization process. We implement our method in consistent with the vision-transformer (ViT) backbone and show that our proposed method can outperform state-of-the-art ViT based UDA methods on diverse benchmarks (DomainNet, Office-Home, and VISDA2017).

This paper proposes a feature synthesis approach for zero-shot semantic segmentation of 3D point clouds, enabling generalization to previously unseen categories. Given only the class-level semantic information for unseen objects, we strive to enhance the correspondence, alignment and consistency between the visual and semantic spaces, to synthesise diverse, generic and transferable visual features. We develop a masked learning strategy to promote diversity within the same class visual features and enhance the separation between different classes. We further cast the visual features into a prototypical space to model their distribution for alignment with the corresponding semantic space. Finally, we develop a consistency regularizer to preserve the semantic-visual relationships between the realseen features and synthetic-unseen features. Our approach shows considerable semantic segmentation gains on ScanNet, S3DIS and SemanticKITTI benchmarks. Our code is available at: https://github.com/leolyj/3DPC-GZSL

Knowledge distillation efficiently improves a small model’s performance by mimicking the teacher model’s behavior. Most existing methods assume that distilling from a large and accurate teacher model leads to better student models. However, several studies show the difficulty of distillation from large teacher models and opt for heuristics to address this. In this work, we demonstrate that large teacher models can still be effective in knowledge distillation. We show that the spurious features learned by large models are the cause of difficulty in knowledge distillation for small students. To overcome this issue, we propose employing ℓ1regularization to prevent teacher models from learning an excessive number of spurious features. Our method alleviates the poor learning for small students when there is a significant disparity in size between the teachers and students. We achieve substantial improvement on various architectures, e.g., ResNet, WideResNet, and VGG, on diverse datasets, including CIFAR-100, Tiny-ImageNet, and ImageNet.

Social group detection is a crucial aspect of various robotic applications, including robot navigation and human-robot interactions. To date, a range of model-based techniques have been employed to address this challenge, such as the F-formation and trajectory similarity frameworks. However, these approaches often fail to provide reliable results in crowded and dynamic scenarios. Recent advancements in this area have mainly focused on learning-based methods, such as deep neural networks that use visual content or human pose. Although visual content based methods have demonstrated promising performance on large-scale datasets, their computational complexity poses a significant barrier to their practical use in real-time applications. To address these issues, we propose a simple and efficient framework for social group detection. Our approach explores the impact of motion trajectory on social grouping and utilizes a novel, reliable, and fast data-driven method. We formulate the individuals in a scene as a graph, where the nodes are represented by LSTM-encoded trajectories and the edges are defined by the distances between each pair of tracks. Our framework employs a modified graph transformer module and graph clustering losses to detect social groups. Our experiments on the popular JRDB-Act dataset reveal noticeable improvements in performance, with relative improvements ranging from 2% to 11%. Furthermore, our framework is significantly faster, with up to 12x faster inference times compared to state-of-the-art methods under the same computation resources. These results demonstrate that our proposed method is suitable for real-time robotic applications. .

In recent years, open-vocabulary (OV) object detection has attracted increasing research attention. Unlike traditional detection, which only recognizes fixed-category objects, OV detection aims to detect objects in an open category set. Previous works often leverage vision-language (VL) training data (e.g., referring grounding data) to recognize OV objects. However, they only use pairs of nouns and individual objects in VL data, while these data usually contain much more information, such as scene graphs, which are also crucial for OV detection. In this paper, we propose a novel Scene-Graph-Based Discovery Network (SGDN) that exploits scene graph cues for OV detection. Firstly, a scene-graph-based decoder (SGDecoder) including sparse scene-graph-guided attention (SSGA) is presented. It captures scene graphs and leverages them to discover OV objects. Secondly, we propose scene-graph-based prediction (SGPred), where we build a scene-graph-based offset regression (SGOR) mechanism to enable mutual enhancement between scene graph extraction and object localization. Thirdly, we design a cross-modal learning mechanism in SGPred. It takes scene graphs as bridges to improve the consistency between cross-modal embeddings for OV object classification. Experiments on COCO and LVIS demonstrate the effectiveness of our approach. Moreover, we show the ability of our model for OV scene graph detection, while previous OV scene graph generation methods cannot tackle this task.

We present LAVA, a simple yet effective method for multi-domain visual transfer learning with limited data. LAVA builds on a few recent innovations to enable adapting to partially labelled datasets with class and domain shifts. First, LAVA learns self-supervised visual representations on the source dataset and ground them using class label semantics to overcome transfer collapse problems associated with supervised pretraining. Secondly, LAVA maximises the gains from unlabelled target data via a novel method which uses multi-crop augmentations to obtain highly robust pseudo-labels. By combining these ingredients, LAVA achieves a new state-of-the-art on ImageNet semi-supervised protocol, as well as on 7 out of 10 datasets in multi-domain few-shot learning on the Meta-dataset.1

Most deepfake detection methods focus on detecting spatial and/or spatio-temporal changes in facial attributes and are centered around the binary classification task of detecting whether a video is real or fake. This is because available benchmark datasets contain mostly visual-only modifications present in the entirety of the video. However, a sophisticated deepfake may include small segments of audio or audio-visual manipulations that can completely change the meaning of the video content. To addresses this gap, we propose and benchmark a new dataset, Localized Audio Visual DeepFake (LAV-DF), consisting of strategic content-driven audio, visual and audio-visual manipulations. The proposed baseline method, Boundary Aware Temporal Forgery Detection (BA-TFD), is a 3D Convolutional Neural Network-based architecture which effectively captures multimodal manipulations. We further improve (i.e. BA-TFD+) the baseline method by replacing the backbone with a Multiscale Vision Transformer and guide the training process with contrastive, frame classification, boundary matching and multimodal boundary matching loss functions. The quantitative analysis demonstrates the superiority of BA-TFD+ on temporal forgery localization and deepfake detection tasks using several benchmark datasets including our newly proposed dataset. The dataset, models and code are available at https://github.com/ControlNet/LAV-DF.

Deep Convolution Neural Networks (CNNs) can easily be fooled by subtle, imperceptible changes to the input images. To address this vulnerability, adversarial training creates perturbation patterns and includes them in the training set to robustify the model. In contrast to existing adversarial training methods that only use class-boundary information (e.g., using a cross-entropy loss), we propose to exploit additional information from the feature space to craft stronger adversaries that are in turn used to learn a robust model. Specifically, we use the style and content information of the target sample from another class, alongside its class-boundary information to create adversarial perturbations. We apply our proposed multi-task objective in a deeply supervised manner, extracting multi-scale feature knowledge to create maximally separating adversaries. Subsequently, we propose a max-margin adversarial training approach that minimizes the distance between source image and its adversary and maximizes the distance between the adversary and the target image. Our adversarial training approach demonstrates strong robustness compared to state-of-the-art defenses, generalizes well to naturally occurring corruptions and data distributional shifts, and retains the model's accuracy on clean examples.

Given a degraded input image, image restoration aims to recover the missing high-quality image content. Numerous applications demand effective image restoration, e.g., computational photography, surveillance, autonomous vehicles, and remote sensing. Significant advances in image restoration have been made in recent years, dominated by convolutional neural networks (CNNs). The widely-used CNN-based methods typically operate either on full-resolution or on progressively low-resolution representations. In the former case, spatial details are preserved but the contextual information cannot be precisely encoded. In the latter case, generated outputs are semantically reliable but spatially less accurate. This paper presents a new architecture with a holistic goal of maintaining spatially-precise high-resolution representations through the entire network, and receiving complementary contextual information from the low-resolution representations. The core of our approach is a multi-scale residual block containing the following key elements: (a) parallel multi-resolution convolution streams for extracting multi-scale features, (b) information exchange across the multi-resolution streams, (c) non-local attention mechanism for capturing contextual information, and (d) attention based multi-scale feature aggregation. Our approach learns an enriched set of features that combines contextual information from multiple scales, while simultaneously preserving the high-resolution spatial details. Extensive experiments on six real image benchmark datasets demonstrate that our method, named as MIRNet-v2, achieves state-of-the-art results for a variety of image processing tasks, including defocus deblurring, image denoising, super-resolution, and image enhancement. The source code and pre-trained models are available at https://github.com/swz30/MIRNetv2.

We propose a weakly supervised approach for salient object detection from multi-modal RGB-D data. Our approach only relies on labels from scribbles, which are much easier to annotate, compared with dense labels used in conventional fully supervised setting. In contrast to existing methods that employ supervision signals on the output space, our design regularizes the intermediate latent space to enhance discrimination between salient and non-salient objects. We further introduce a contour detection branch to implicitly constrain the semantic boundaries and achieve precise edges of detected salient objects. To enhance the long-range dependencies among local features, we introduce a Cross-Padding Attention Block (CPAB). Extensive experiments on seven benchmark datasets demonstrate that our method not only outperforms existing weakly supervised methods, but is also on par with several fully-supervised state-of-the-art models. Code is available at https://github.com/leolyj/DHFR-SOD.

Point cloud scene flow estimation is of practical importance for dynamic scene navigation in autonomous driving. Since scene flow labels are hard to obtain, current methods train their models on synthetic data and transfer them to real scenes. However, large disparities between existing synthetic datasets and real scenes lead to poor model transfer. We make two major contributions to address that. First, we develop a point cloud collector and scene flow annotator for GTA-V engine to automatically obtain diverse realistic training samples without human intervention. With that, we develop a large-scale synthetic scene flow dataset GTA-SF. Second, we propose a mean-teacher-based domain adaptation framework that leverages self-generated pseudo-labels of the target domain. It also explicitly incorporates shape deformation regularization and surface correspondence refinement to address distortions and misalignments in domain transfer. Through extensive experiments, we show that our GTA-SF dataset leads to a consistent boost in model generalization to three real datasets (i.e., Waymo, Lyft and KITTI) as compared to the most widely used FT3D dataset. Moreover, our framework achieves superior adaptation performance on six source-target dataset pairs, remarkably closing the average domain gap by 60%. Data and codes are available at https://github.com/leolyj/DCA-SRSFE

Active learning (AL) is a promising ML paradigm that has the potential to parse through large unlabeled data and help reduce annotation cost in domains where labeling data can be prohibitive. Recently proposed neural network based AL methods use different heuristics to accomplish this goal. In this study, we demonstrate that under identical experimental settings, different types of AL algorithms (uncertainty based, diversity based, and committee based) produce an inconsistent gain over random sampling baseline. Through a variety of experiments, controlling for sources of stochasticity, we show that variance in performance metrics achieved by AL algorithms can lead to results that are not consistent with the previously reported results. We also found that under strong regularization, AL methods show marginal or no advantage over the random sampling baseline under a variety of experimental conditions. Finally, we conclude with a set of recommendations on how to assess the results using a new AL algorithm to ensure results are reproducible and robust under changes in experimental conditions. We share our codes to facilitate AL evaluations. We believe our findings and recommendations will help advance reproducible research in AL using neural networks.

Deep models trained on source domain lack generalization when evaluated on unseen target domains with different data distributions. The problem becomes even more pro-nounced when we have no access to target domain samples for adaptation. In this paper, we address domain generalized semantic segmentation, where a segmentation model is trained to be domain-invariant without using any target domain data. Existing approaches to tackle this problem standardize data into a unified distribution. We argue that while such a standardization promotes global normalization, the resulting features are not discriminative enough to get clear segmentation boundaries. To enhance separation between categories while simultaneously promoting domain invariance, we propose a framework including two novel modules: Semantic-Aware Normalization (SAN) and Semantic-Aware Whitening (SAW). Specifically, SAN focuses on category-level center alignment between features from different image styles, while SAW enforces distributed alignment for the already center-aligned features. With the help of SAN and SAW, we encourage both intra-category compactness and inter-category separability. We validate our approach through extensive experiments on widely-used datasets (i.e. GTAV, SYNTHIA, Cityscapes, Mapillary and BDDS). Our approach shows significant improvements over existing state-of-the-art on various backbone networks. Code is available at https://github.com/leolyj/SAN-SAW

Object proposal generation is an important and fundamental task in computer vision. In this paper, we propose ProposalCLIP, a method towards unsupervised open-category object proposal generation. Unlike previous works which require a large number of bounding box annotations and/or can only generate proposals for limited object categories, our ProposalCLIP is able to predict proposals for a large variety of object categories without annotations, by exploiting CLIP (contrastive language-image pre-training) cues. Firstly, we analyze CLIP for unsupervised open-category proposal generation and design an objectness score based on our empirical analysis on proposal selection. Secondly, a graph-based merging module is proposed to solve the limitations of CLIP cues and merge fragmented proposals. Finally, we present a proposal regression module that extracts pseudo labels based on CLIP cues and trains a lightweight network to further refine proposals. Extensive experiments on PASCAL VOC, COCO and Visual Genome datasets show that our ProposalCLIP can better generate proposals than previous state-of-the-art methods. Our ProposalCLIP also shows benefits for downstream tasks, such as unsupervised object detection.

Since convolutional neural networks (CNNs) perform well at learning generalizable image priors from large-scale data, these models have been extensively applied to image restoration and related tasks. Recently, another class of neural architectures, Transformers, have shown significant performance gains on natural language and high-level vision tasks. While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images. In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images. Our model, named Restoration Transformer (Restormer), achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising). The source code and pre-trained models are available at https://github.com/swz30/Restormer.

In challenging real-life conditions such as extreme head-pose, occlusions, and low-resolution images where the visual information fails to estimate visual attention/gaze direction, audio signals could provide important and complementary information. In this paper, we explore if audio-guided coarse head-pose can further enhance visual attention estimation performance for non-prolific faces. Since it is difficult to annotate audio signals for estimating the head-pose of the speaker, we use off-the-shelf state-of-the-art models to facilitate cross-modal weak-supervision. During the training phase, the framework learns complementary information from synchronized audio-visual modality. Our model can utilize any of the available modalities i.e. audio, visual or audio-visual for task-specific inference. It is interesting to note that, when AV-Gaze is tested on benchmark datasets with these specific modalities, it achieves competitive results on multiple datasets, while being highly adaptive toward challenging scenarios.

Robust gaze estimation is a challenging task, even for deep CNNs, due to the non-availability of large-scale labeled data. Moreover, gaze annotation is a time-consuming process and requires specialized hardware setups. We propose MTGLS: a Multi-Task Gaze estimation framework with Limited Supervision, which leverages abundantly available non-annotated facial image data. MTGLS distills knowledge from off-the-shelf facial image analysis models, and learns strong feature representations of human eyes, guided by three complementary auxiliary signals: (a) the line of sight of the pupil (i.e. pseudo-gaze) defined by the localized facial landmarks, (b) the head-pose given by Euler angles, and (c) the orientation of the eye patch (left/right eye). To overcome inherent noise in the supervisory signals, MT-GLS further incorporates a noise distribution modelling approach. Our experimental results show that MTGLS learns highly generalized representations which consistently perform well on a range of datasets. Our proposed framework outperforms the unsupervised state-of-the-art on CAVE (by ∼ 6.43%) and even supervised state-of-the-art methods on Gaze360 (by ∼ 6.59%) datasets.

Image restoration tasks demand a complex balance between spatial details and high-level contextualized information while recovering images. In this paper, we propose a novel synergistic design that can optimally balance these competing goals. Our main proposal is a multi-stage architecture, that progressively learns restoration functions for the degraded inputs, thereby breaking down the overall recovery process into more manageable steps. Specifically, our model first learns the contextualized features using encoder-decoder architectures and later combines them with a high-resolution branch that retains local information. At each stage, we introduce a novel per-pixel adaptive design that leverages in-situ supervised attention to reweight the local features. A key ingredient in such a multi-stage architecture is the information exchange between different stages. To this end, we propose a two-faceted approach where the information is not only exchanged sequentially from early to late stages, but lateral connections between feature processing blocks also exist to avoid any loss of information. The resulting tightly interlinked multi-stage architecture, named as MPRNet, delivers strong performance gains on ten datasets across a range of tasks including image deraining, deblurring, and denoising. The source code and pre-trained models are available at https://github.com/swz30/MPRNet.

This work proposes a weakly-supervised temporal action localization framework, called D2-Net, which strives to temporally localize actions using video-level supervision. Our main contribution is the introduction of a novel loss formulation, which jointly enhances the discriminability of latent embeddings and robustness of the output temporal class activations with respect to foreground-background noise caused by weak supervision. The proposed formulation comprises a discriminative and a denoising loss term for enhancing temporal action localization. The discriminative term incorporates a classification loss and utilizes a top-down attention mechanism to enhance the separability of latent foreground-background embeddings. The denoising loss term explicitly addresses the foreground-background noise in class activations by simultaneously maximizing intra-video and inter-video mutual information using a bottom-up attention mechanism. As a result, activations in the foreground regions are emphasized whereas those in the background regions are suppressed, thereby leading to more robust predictions. Comprehensive experiments are performed on multiple benchmarks, including THUMOS14 and ActivityNet1.2. Our D2-Net performs favorably in comparison to the existing methods on all datasets, achieving gains as high as 2.3% in terms of mAP at IoU=0.5 on THUMOS14. Source code is available at https://github.com/naraysa/D2-Net.

While the untargeted black-box transferability of adversarial perturbations has been extensively studied before, changing an unseen model’s decisions to a specific ‘targeted’ class remains a challenging feat. In this paper, we propose a new generative approach for highly transferable targeted perturbations (TTP). We note that the existing methods are less suitable for this task due to their reliance on class-boundary information that changes from one model to another, thus reducing transferability. In contrast, our approach matches the perturbed image ‘distribution’ with that of the target class, leading to high targeted transferability rates. To this end, we propose a new objective function that not only aligns the global distributions of source and target images, but also matches the local neighbourhood structure between the two domains. Based on the proposed objective, we train a generator function that can adaptively synthesize perturbations specific to a given input. Our generative approach is in-dependent of the source or target domain labels, while consistently performs well against state-of-the-art methods on a wide range of attack settings. As an example, we achieve 32.63% target transferability from (an adversarially weak) VGG19BNto (a strong) WideResNet on ImageNet val. set, which is 4× higher than the previous best generative attack and 16× better than instance-specific iterative attack. Code is available at: https://github.com/Muzammal-Naseer/TTP.

Deep neural networks have achieved remarkable performance on a range of classification tasks, with softmax crossentropy (CE) loss emerging as the de-facto objective function. The CE loss encourages features of a class to have a higher projection score on the true class-vector compared to the negative classes. However, this is a relative constraint and does not explicitly force different class features to be well-separated. Motivated by the observation that ground-truth class representations in CE loss are orthogonal (one-hot encoded vectors), we develop a novel loss function termed ‘Orthogonal Projection Loss' (OPL) which imposes orthogonality in the feature space. OPL augments the properties of CE loss and directly enforces inter-class separation alongside intra-class clustering in the feature space through orthogonality constraints on the mini-batch level. As compared to other alternatives of CE, OPL offers unique advantages e.g., no additional learnable parameters, does not require careful negative mining and is not sensitive to the batch size. Given the plug-and-play nature of OPL, we evaluate it on a diverse range of tasks including image recognition (CIFAR-100), large-scale classification (ImageNet), domain generalization (PACS) and few-shot learning (mini-ImageNet, CIFAR-FS, tiered-ImageNet and Meta-dataset) and demonstrate its effectiveness across the board. Furthermore, OPL offers better robustness against practical nuisances such as adversarial attacks and label noise. Code is available at: https://github.com/kahnchana/opl.

Vision transformers (ViT) have demonstrated impressive performance across numerous machine vision tasks. These models are based on multi-head self-attention mechanisms that can flexibly attend to a sequence of image patches to encode contextual cues. An important question is how such flexibility (in attending image-wide context conditioned on a given patch) can facilitate handling nuisances in natural images e.g., severe occlusions, domain shifts, spatial permutations, adversarial and natural perturbations. We systematically study this question via an extensive set of experiments encompassing three ViT families and provide comparisons with a high-performing convolutional neural network (CNN). We show and analyze the following intriguing properties of ViT: (a)Transformers are highly robust to severe occlusions, perturbations and domain shifts, e.g., retain as high as 60% top-1 accuracy on ImageNet even after randomly occluding 80% of the image content. (b)The robustness towards occlusions is not due to texture bias, instead we show that ViTs are significantly less biased towards local textures, compared to CNNs. When properly trained to encode shape-based features, ViTs demonstrate shape recognition capability comparable to that of human visual system, previously unmatched in the literature. (c)Using ViTs to encode shape representation leads to an interesting consequence of accurate semantic segmentation without pixel-level supervision. (d)Off-the-shelf features from a single ViT model can be combined to create a feature ensemble, leading to high accuracy rates across a range of classification datasets in both traditional and few-shot learning paradigms. We show effective features of ViTs are due to flexible and dynamic receptive fields possible via self-attention mechanisms. Our code will be publicly released.

Deep neural networks can easily be fooled by an adversary with minuscule perturbations added to an input image. The existing defense techniques suffer greatly under white-box attack settings, where an adversary has full knowledge of the network and can iterate several times to find strong perturbations. We observe that the main reason for the existence of such vulnerabilities is the close proximity of different class samples in the learned feature space of deep models. This allows the model decisions to be completely changed by adding an imperceptible perturbation to the inputs. To counter this, we propose to class-wise disentangle the intermediate feature representations of deep networks, specifically forcing the features for each class to lie inside a convex polytope that is maximally separated from the polytopes of other classes. In this manner, the network is forced to learn distinct and distant decision regions for each class. We observe that this simple constraint on the features greatly enhances the robustness of learned models, even against the strongest white-box attacks, without degrading the classification performance on clean images. We report extensive evaluations in both black-box and white-box attack scenarios and show significant gains in comparison to state-of-the-art defenses.

Adversarial examples can cause catastrophic mistakes in Deep Neural Network (DNNs) based vision systems e.g., for classification, segmentation and object detection. The vulnerability of DNNs against such attacks can prove a major roadblock towards their real-world deployment. Transferability of adversarial examples demand generalizable defenses that can provide cross-task protection. Adversarial training that enhances robustness by modifying target model's parameters lacks such generalizability. On the other hand, different input processing based defenses fall short in the face of continuously evolving attacks. In this paper, we take the first step to combine the benefits of both approaches and propose a self-supervised adversarial training mechanism in the input space. By design, our defense is a generalizable approach and provides significant robustness against the unseen adversarial attacks (e.g. by reducing the success rate of translation-invariant ensemble attack from 82.6% to 31.9% in comparison to previous stateof-the-art). It can be deployed as a plug-and-play solution to protect a variety of vision systems, as we demonstrate for the case of classification, segmentation and detection. Code is available at: https://github.com/ Muzammal-Naseer/NRP.

Humans can continuously learn new knowledge as their experience grows. In contrast, previous learning in deep neural networks can quickly fade out when they are trained on a new task. In this paper, we hypothesize this problem can be avoided by learning a set of generalized parameters, that are neither specific to old nor new tasks. In this pursuit, we introduce a novel meta-learning approach that seeks to maintain an equilibrium between all the encountered tasks. This is ensured by a new meta-update rule which avoids catastrophic forgetting. In comparison to previous meta-learning techniques, our approach is task-agnostic. When presented with a continuum of data, our model automatically identifies the task and quickly adapts to it with just a single update. We perform extensive experiments on five datasets in a class-incremental setting, leading to significant improvements over the state of the art methods (e.g., a 21.3% boost on CIFAR100 with 10 incremental tasks). Specifically, on large-scale datasets that generally prove difficult cases for incremental learning, our approach delivers absolute gains as high as 19.1% and 7.4% on ImageNet and MS-Celeb datasets, respectively.

The availability of large-scale datasets has helped unleash the true potential of deep convolutional neural networks (CNNs). However, for the single-image denoising problem, capturing a real dataset is an unacceptably expensive and cumbersome procedure. Consequently, image denoising algorithms are mostly developed and evaluated on synthetic data that is usually generated with a widespread assumption of additive white Gaussian noise (AWGN). While the CNNs achieve impressive results on these synthetic datasets, they do not perform well when applied on real camera images, as reported in recent benchmark datasets. This is mainly because the AWGN is not adequate for modeling the real camera noise which is signal-dependent and heavily transformed by the camera imaging pipeline. In this paper, we present a framework that models camera imaging pipeline in forward and reverse directions. It allows us to produce any number of realistic image pairs for denoising both in RAW and sRGB spaces. By training a new image denoising network on realistic synthetic data, we achieve the state-of-the-art performance on real camera benchmark datasets. The parameters in our models are ~5 times lesser than the previous best method for RAW denoising. Furthermore, we demonstrate that the proposed framework generalizes beyond image denoising problem e.g., for color matching in stereoscopic cinema. The source code and pre-trained models are available at https://github.com/swz30/CycleISP.

In this paper, we propose a multi-modal mesh surface representation by fusing texture and geometric data. Our approach defines an inverse mapping between different geometric descriptors computed on the mesh surface, and the corresponding 2D texture image of the mesh, allowing the construction of fused geometrically augmented images. This new fused modality enables us to learn feature representations from 3D data in a highly efficient manner by employing standard convolutional neural networks in a transfer-learning mode. In contrast to existing methods, the proposed approach is both computationally and memory efficient, preserves intrinsic geometric information and learns highly discriminative feature representations by effectively fusing shape and texture information at the data level. The efficacy is demonstrated on the task of facial expression classification, showing competitive performance with state-of-the-art methods.

In this paper, we propose an approach to learn generic multi-modal mesh surface representations using a novel scheme for fusing texture and geometric data. Our approach defines an inverse mapping between different geometric descriptors computed on the mesh surface or its down-sampled version, and the corresponding 2D texture image of the mesh, allowing the construction of fused geometrically augmented images (FGAI). This new fused modality enables us to learn feature representations from 3D data in a highly efficient manner by simply employing standard CNNs in a transfer-learning mode. The proposed approach is both computationally and memory efficient, preserves intrinsic geometric information and learns highly discriminative feature representations by effectively fusing shape and texture information at data level. The efficacy of our approach is demonstrated for the tasks of facial action unit detection and expression classification. The extensive experiments conducted on the Bosphorus and BU-4DFE datasets show that our method produces a significant boost in the performance when compared to state-of-the-art solutions.

Convolutional Neural Networks have achieved significant success across multiple computer vision tasks. However, they are vulnerable to carefully crafted, human-imperceptible adversarial noise patterns which constrain their deployment in critical security-sensitive systems. This paper proposes a computationally efficient image enhancement approach that provides a strong defense mechanism to effectively mitigate the effect of such adversarial perturbations. We show that deep image restoration networks learn mapping functions that can bring off-the-manifold adversarial samples onto the natural image manifold, thus restoring classification towards correct classes. A distinguishing feature of our approach is that, in addition to providing robustness against attacks, it simultaneously enhances image quality and retains models performance on clean images. Furthermore, the proposed method does not modify the classifier or requires a separate mechanism to detect adversarial images. The effectiveness of the scheme has been demonstrated through extensive experiments, where it has proven a strong defense in gray-box settings. The proposed scheme is simple and has the following advantages: (1) it does not require any model training or parameter optimization, (2) it complements other existing defense mechanisms, (3) it is agnostic to the attacked model and attack type and (4) it provides superior performance across all popular attack algorithms. Our codes are publicly available at https://github.com/aamir-mustafa/super-resolution-adversarial-defense.

Learning unbiased models on imbalanced datasets is a significant challenge. Rare classes tend to get a concentrated representation in the classification space which hampers the generalization of learned boundaries to new test examples. In this paper, we demonstrate that the Bayesian uncertainty estimates directly correlate with the rarity of classes and the difficulty level of individual samples. Subsequently, we present a novel framework for uncertainty based class imbalance learning that follows two key insights: First, classification boundaries should be extended further away from a more uncertain (rare) class to avoid over-fitting and enhance its generalization. Second, each sample should be modeled as a multi-variate Gaussian distribution with a mean vector and a covariance matrix defined by the sample's uncertainty. The learned boundaries should respect not only the individual samples but also their distribution in the feature space. Our proposed approach efficiently utilizes sample and class uncertainty information to learn robust features and more generalizable classifiers. We systematically study the class imbalance problem and derive a novel loss formulation for max-margin learning based on Bayesian uncertainty measure. The proposed method shows significant performance improvements on six benchmark datasets for face verification, attribute prediction, digit/object classification and skin lesion detection.

Real-world object classes appear in imbalanced ratios. This poses a significant challenge for classifiers which get biased towards frequent classes. We hypothesize that improving the generalization capability of a classifier should improve learning on imbalanced datasets. Here, we introduce the first hybrid loss function that jointly performs classification and clustering in a single formulation. Our approach is based on an 'affinity measure' in Euclidean space that leads to the following benefits: (1) direct enforcement of maximum margin constraints on classification boundaries, (2) a tractable way to ensure uniformly spaced and equidistant cluster centers, (3) flexibility to learn multiple class prototypes to support diversity and discriminability in feature space. Our extensive experiments demonstrate the significant performance improvements on visual classification and verification tasks on multiple imbalanced datasets. The proposed loss can easily be plugged in any deep architecture as a differentiable block and demonstrates robustness against different levels of data imbalance and corrupted labels.

Deep neural networks are vulnerable to adversarial attacks which can fool them by adding minuscule perturbations to the input images. The robustness of existing defenses suffers greatly under white-box attack settings, where an adversary has full knowledge about the network and can iterate several times to find strong perturbations. We observe that the main reason for the existence of such perturbations is the close proximity of different class samples in the learned feature space. This allows model decisions to be totally changed by adding an imperceptible perturbation in the inputs. To counter this, we propose to class-wise disentangle the intermediate feature representations of deep networks. Specifically, we force the features for each class to lie inside a convex polytope that is maximally separated from the polytopes of other classes. In this manner, the network is forced to learn distinct and distant decision regions for each class. We observe that this simple constraint on the features greatly enhances the robustness of learned models, even against the strongest white-box attacks, without degrading the classification performance on clean images. We report extensive evaluations in both black-box and white-box attack scenarios and show significant gains in comparison to state-of-the art defenses.

Incremental life-long learning is a main challenge towards the long-standing goal of Artificial General Intelligence. In real-life settings, learning tasks arrive in a sequence and machine learning models must continually learn to increment already acquired knowledge. The existing incremental learning approaches fall well below the state-of-the-art cumulative models that use all training classes at once. In this paper, we propose a random path selection algorithm, called RPS-Net, that progressively chooses optimal paths for the new tasks while encouraging parameter sharing and reuse. Our approach avoids the overhead introduced by computationally expensive evolutionary and reinforcement learning based path selection strategies while achieving considerable performance gains. As an added novelty, the proposed model integrates knowledge distillation and retrospection along with the path selection strategy to overcome catastrophic forgetting. In order to maintain an equilibrium between previous and newly acquired knowledge, we propose a simple controller to dynamically balance the model plasticity. Through extensive experiments, we demonstrate that the proposed method surpasses the state-of-the-art performance on incremental learning and by utilizing parallel computation this method can run in constant time with nearly the same efficiency as a conventional deep convolutional neural network.

Variational auto-encoders (VAEs) provide an attractive solution to image generation problem. However, they tend to produce blurred and over-smoothed images due to their dependence on pixel-wise reconstruction loss. This paper introduces a new approach to alleviate this problem in the VAE based generative models. Our model simultaneously learns to match the data, reconstruction loss and the latent distributions of real and fake images to improve the quality of generated samples. To compute the loss distributions, we introduce an auto-encoder based discriminator model which allows an adversarial learning procedure. The discriminator in our model also provides perceptual guidance to the VAE by matching the learned similarity metric of the real and fake samples in the latent space. To stabilize the overall training process, our model uses an error feedback approach to maintain the equilibrium between competing networks in the model. Our experiments show that the generated samples from our proposed model exhibit a diverse set of attributes and facial expressions and scale up to highresolution images very well.

This note clarifies the experimental settings of [1] and shows that the issue raised by [2] is due to a lack of details in [1] which resulted in a misinterpretation of the experimental settings.

Class imbalance is a common problem in the case of real-world object detection and classification tasks. Data of some classes are abundant, making them an overrepresented majority, and data of other classes are scarce, making them an underrepresented minority. This imbalance makes it challenging for a classifier to appropriately learn the discriminating boundaries of the majority and minority classes. In this paper, we propose a cost-sensitive (CoSen) deep neural network, which can automatically learn robust feature representations for both the majority and minority classes. During training, our learning procedure jointly optimizes the class-dependent costs and the neural network parameters. The proposed approach is applicable to both binary and multiclass problems without any modification. Moreover, as opposed to data-level approaches, we do not alter the original data distribution, which results in a lower computational cost during the training process. We report the results of our experiments on six major image classification data sets and show that the proposed approach significantly outperforms the baseline algorithms. Comparisons with popular data sampling techniques and CoSen classifiers demonstrate the superior performance of our proposed method.

Automated facial video analysis is useful in numerous health care applications. For example, spatio-temporal analysis of such videos has been previously done for assisting clinicians in the diagnosis of depression. Physiological measures, such as an individual's heart rate, provide very important cues to understand a person's mental health. Unobtrusively estimated heart rate has not been previously used to analyse individuals' mental health. In this paper, we automatically estimate heart rate activity from facial videos. We then study the association of the estimated heart rate activity with the person's mental health, as diagnosed by clinicians. Specifically, from the heart rate activity in response to watching different movies, we classify individuals as either depressed or healthy. The efficacy of the proposed scheme is demonstrated by experimental evaluations on a clinically validated dataset. Our results suggest unobtrusively estimated heart rate to be very effective for depression analysis.

Recent advances in deep learning have resulted in human-level performances on popular unconstrained face datasets including Labeled Faces in the Wild and YouTube Faces. To further advance research, IJB-A benchmark was recently introduced with more challenges especially in the form of extreme head poses. Registration of such faces is quite demanding and often requires laborious procedures like facial landmark localization. In this paper, we propose a Convolutional Neural Networks based data-driven approach which learns to simultaneously register and represent faces. We validate the proposed scheme on template based unconstrained face identification. Here, a template contains multiple media in the form of images and video frames. Unlike existing methods which synthesize all template media information at feature level, we propose to keep the template media intact. Instead, we represent gallery templates by their trained one-vs-rest discriminative models and then employ a Bayesian strategy which optimally fuses decisions of all medias in a query template. We demonstrate the efficacy of the proposed scheme on IJB-A, YouTube Celebrities and COX datasets where our approach achieves significant relative performance boosts of 3.6%, 21.6% and 12.8% respectively.

Recent years have seen a lot of activity in affective computing for automated analysis of depression. However, no research has so far directly evaluated the performance of facial behavioural analysis methods in classifying different subtypes of depression such as melancholia. The mental state assessment of a mood disorder depends largely on appearance, behaviour, speech, thought, perception, mood and facial affect. Mood and facial affect mainly contribute to distinguishing melancholia from nonmelancholia. These are assessed by clinicians, and hence vulnerable to subjective judgement. As a result, clinical assessment alone may not accurately capture the presence or absence of specific disorders such as melancholia, a distressing condition whose presence has important treatment implications. Melancholia is characterised by severe anhedonia and psychomotor disturbance, which can be a mix of motor retardation with periods of superimposed agitation. To the best of our knowledge, this study is the first attempt to perform facial behavioural analysis to disambiguate melancholia from non-melancholia and healthy controls on the basis of facial behavioural characteristics. We report the sensitivity and specificity of classification in depressive subtypes. These results serve as a baseline for more fine-grained depression classification and analysis.

Spectral signatures of natural scenes were earlier found to be distinctive for different scene types with varying spatial envelope properties such as openness, naturalness, ruggedness, and symmetry. Recently, such handcrafted features have been outclassed by deep learning based representations. This paper proposes a novel spectral description of convolution features, implemented efficiently as a unitary transformation within deep network architectures. To the best of our knowledge, this is the first attempt to use deep learning based spectral features explicitly for image classification task. We show that the spectral transformation decorrelates convolutional activations, which reduces co-adaptation between feature detections, thus acts as an effective regularizer. Our approach achieves significant improvements on three large-scale scene-centric datasets (MIT-67, SUN-397, and Places-205). Furthermore, we evaluated the proposed approach on the attribute detection task where its superior performance manifests its relevance to semantically meaningful characteristics of natural scenes.

Recent years have seen a lot of activity in affective computing for automated analysis of depression. However, no research has so far proposed a multimodal system for classifying different subtypes of depression such as melancholia. The mental state assessment of a mood disorder depends primarily on appearance, behaviour, speech, thought, perception, mood and facial affect. Mood and facial affect mainly contribute to distinguishing melancholia from non-melancholia. These are assessed by clinicians, and hence vulnerable to subjective judgement. As a result, clinical assessment alone may not accurately capture the presence or absence of specific disorders such as melancholia, a distressing condition whose presence has important treatment implications. Melancholia is characterised by severe anhedonia and psychomotor disturbance, which can be a combination of motor retardation with periods of superimposed agitation. Psychomotor disturbance can be sensed in both face and voice. To the best of our knowledge, this study is the first attempt to propose a multimodal system to differentiate melancholia from non-melancholia and healthy controls. We report the sensitivity and specificity of classification in depressive subtypes.

Unlike standard object classification, where the image to be classified contains one or multiple instances of the same object, indoor scene classification is quite different since the image consists of multiple distinct objects. Furthermore, these objects can be of varying sizes and are present across numerous spatial locations in different layouts. For automatic indoor scene categorization, large-scale spatial layout deformations and scale variations are therefore two major challenges and the design of rich feature descriptors which are robust to these challenges is still an open problem. This paper introduces a new learnable feature descriptor called “spatial layout and scale invariant convolutional activations” to deal with these challenges. For this purpose, a new convolutional neural network architecture is designed which incorporates a novel “spatially unstructured” layer to introduce robustness against spatial layout deformations. To achieve scale invariance, we present a pyramidal image representation. For feasible training of the proposed network for images of indoor scenes, this paper proposes a methodology, which efficiently adapts a trained network model (on a large-scale data) for our task with only a limited amount of available training data. The efficacy of the proposed approach is demonstrated through extensive experiments on a number of data sets, including MIT-67, Scene-15, Sports-8, Graz-02, and NYU data sets.

Indoor scene recognition is a multi-faceted and challenging problem due to the diverse intra-class variations and the confusing inter-class similarities. This paper presents a novel approach which exploits rich mid-level convolutional features to categorize indoor scenes. Traditionally used convolutional features preserve the global spatial structure, which is a desirable property for general object recognition. However, we argue that this structuredness is not much helpful when we have large variations in scene layouts, e.g., in indoor scenes. We propose to transform the structured convolutional activations to another highly discriminative feature space. The representation in the transformed space not only incorporates the discriminative aspects of the target dataset, but it also encodes the features in terms of the general object categories that are present in indoor scenes. To this end, we introduce a new large-scale dataset of 1300 object categories which are commonly present in indoor scenes. Our proposed approach achieves a significant performance boost over previous state of the art approaches on five major scene classification datasets.

To find the optimal nonlinear separating boundary with maximum margin in the input data space, this paper proposes Contractive Rectifier Networks (CRNs), wherein the hidden-layer transformations are restricted to be contraction mappings. The contractive constraints ensure that the achieved separating margin in the input space is larger than or equal to the separating margin in the output layer. The training of the proposed CRNs is formulated as a linear support vector machine (SVM) in the output layer, combined with two or more contractive hidden layers. Effective algorithms have been proposed to address the optimization challenges arising from contraction constraints. Experimental results on MNIST, CIFAR-10, CIFAR-100 and MIT-67 datasets demonstrate that the proposed contractive rectifier networks consistently outperform their conventional unconstrained rectifier network counterparts.

Image set classification finds its applications in a number of real-life scenarios such as classification from surveillance videos, multi-view camera networks and personal albums. Compared with single image based classification, it offers more promises and has therefore attracted significant research attention in recent years. Unlike many existing methods which assume images of a set to lie on a certain geometric surface, this paper introduces a deep learning framework which makes no such prior assumptions and can automatically discover the underlying geometric structure. Specifically, a Template Deep Reconstruction Model (TDRM) is defined whose parameters are initialized by performing unsupervised pre-training in a layer-wise fashion using Gaussian Restricted Boltzmann Machines (GRBMs). The initialized TDRM is then separately trained for images of each class and class-specific DRMs are learnt. Based on the minimum reconstruction errors from the learnt class-specific models, three different voting strategies are devised for classification. Extensive experiments are performed to demonstrate the efficacy of the proposed framework for the tasks of face and object recognition from image sets. Experimental results show that the proposed method consistently outperforms the existing state of the art methods.

We propose a deep learning framework for image set classification with application to face recognition. An Adaptive Deep Network Template (ADNT) is defined whose parameters are initialized by performing unsupervised pre-training in a layer-wise fashion using Gaussian Restricted Boltzmann Machines (GRBMs). The pre-initialized ADNT is then separately trained for images of each class and class-specific models are learnt. Based on the minimum reconstruction error from the learnt class-specific models, a majority voting strategy is used for classification. The proposed framework is extensively evaluated for the task of image set classification based face recognition on Honda/UCSD, CMU Mobo, YouTube Celebrities and a Kinect dataset. Our experimental results and comparisons with existing state-of-the-art methods show that the proposed method consistently achieves the best performance on all these datasets.

Most of the existing research on 3D facial expression recognition has been done using static 3D meshes. 3D videos of a face are believed to contain more information in terms of the facial dynamics which are very critical for expression recognition. This paper presents a fully automatic framework which exploits the dynamics of textured 3D videos for recognition of six discrete facial expressions. Local video-patches of variable lengths are extracted from numerous locations of the training videos and represented as points on the Grassmannian manifold. An efficient graph-based spectral clustering algorithm is used to separately cluster these points for every expression class. Using a valid Grassmannian kernel function, the resulting cluster centers are embedded into a Reproducing Kernel Hilbert Space (RKHS) where six binary SVM models are learnt. Given a query video, we extract video-patches from it, represent them as points on the manifold and match these points with the learnt SVM models followed by a voting based strategy to decide about the class of the query video. The proposed framework is also implemented in parallel on 2D videos and a score level fusion of 2D & 3D videos is performed for performance improvement of the system. The experimental results on BU4DFE data set show that the system achieves a very high classification accuracy for facial expression recognition from 3D videos.

This paper presents a fully automatic system which exploits the dynamics of 3D videos and is capable of recognizing six basic facial expressions. Local video-patches of variable lengths are extracted from different locations of the training videos and represented as points on the Grass-mannian manifold. An efficient spectral clustering based algorithm is used to separately cluster points for each of the six expression classes. The resulting cluster centers are matched with the points of a test video and a voting based strategy is used to decide about the expression class of the test video. The proposed system is tested on the largest publicly available 3D video database, BU4DFE. The experimental results show that the system achieves a very high classification accuracy and outperforms the current state of the art algorithms for facial expression recognition from 3D videos.

Local spatiotemporal detectors and descriptors have recently become very popular for video analysis in many applications. They do not require any preprocessing steps and are invariant to spatial and temporal scales. Despite their computational simplicity, they have not been evaluated and tested for video analysis of facial data. This paper considers two space-time detectors and four descriptors and uses bag of features framework for human facial expression recognition on BU_4DFE data set. A comparison of local spatiotemporal features with other non-spatiotemporal published techniques on the same data set is also given. Unlike spatiotemporal features, these techniques involve time consuming and computationally intensive preprocessing steps like manual initialization and tracking of facial points. Our results show that despite being totally automatic and not requiring any user intervention, local spacetime features provide promising and comparable performance for facial expression recognition on BU_4DFE data set.

In this paper, we present a system based on novel low level local features to recognize 3D faces under varying facial expressions. Our local features are obtained by combinatorially selecting two points from expression insensitive semi-rigid portions of the face. The curve length between the two points is computed and the distribution of such curve lengths is used as a feature vector to model the geometric shape distribution of the face. Our proposed features are very simple to compute yet highly distinctive and discriminating. Kernel Fisher discriminant analysis is used for feature optimization, followed by a linear support vector machine classifier for recognition. The system is extensively tested on 2500 facial scans of BU 3DFE dataset. Our experimental results show that the proposed system achieves a very high average classification rate of 99.17% and verification rates of 99.0% and above for a false acceptance rate of 0.001.