Efficiently and accurately recognizing interesting objects within the image and regressing bounding boxes to enclose them has been a persistent pursuit in object detection. However, existing detectors fail to achieve both aspects simultaneously due to insufficient task interaction and suboptimal classification behavior. To solve the problem, this paper proposes a novel detector with Efficient Asymmetric Progressive Semi-Decoupled Head (EAPSDH) and Harmonic Focal Loss (HFL). Specifically, we generalize the detection head into a progressive asymmetric paradigm that performs hierarchical and dynamically recalibrated interaction between classification and localization, enabling iterative mutual enhancement in an efficient manner beyond the prior designs. Meanwhile, HFL is proposed to improve classifier optimization by addressing the imbalance between positive and negative samples. HFL dynamically increases the loss weights of positive samples, amplifying their gradient contributions during classifier training, which significantly reduces classification error. By jointly improving task-specific feature representation and classification optimization, EAPSDH and HFL complement each other to alleviate the inconsistency between classification and localization performance, resulting in an efficient and accurate one-stage detector termed EADet. Experimental results on the MS COCO database demonstrate that EADet effectively mitigates the inconsistency between classification and localization performance. Furthermore, EADet achieves a strong trade-off between accuracy and speed, reaching 47.4 AP at 33.2 FPS on the MS COCO with ResNet-101 under the $2\times $ training schedule, demonstrating its effectiveness compared with recent state-of-the-art detectors. Code will be available at https://github.com/HB-X/EADet.

Existing object detection methods struggle to generalize across increasingly data domains while simultaneously adapting to the emergence of novel categories. To tackle this challenge, adaptive open-set object detection (AOOD) has been introduced, which employs supervised training on base categories within the source domain while enabling unsupervised adaptation to both base and novel categories in the target domain. However, existing AOOD approaches are still hindered by several limitations, including insufficient cross-domain feature representation, inter-category ambiguity in novel classes, and inherent feature bias toward the source domain. To overcome these issues, this paper proposes a category-level collaboration knowledge mining strategy designed to comprehensively exploit both inter-class and intra-class feature relationships across domains. Specifically, a clustering-based memory bank (CMB) is initially constructed to aggregate class prototype features, class auxiliary features, and intra-class disparity features, thereby embedding rich category-level knowledge into a unified memory structure. The CMB is iteratively updated through unsupervised clustering, which facilitates the modeling of intra-category relationships and enhances its capacity for cross-domain knowledge representation. Subsequently, a base-to-novel selection metric (BNSM) is designed to identify features corresponding to novel categories within the source domain by regulating the relationships between the novel categories and each base category. The selected features are then leveraged to initialize the object detector for the classification of novel categories. Finally, an adaptive feature assignment (AFA) strategy is introduced to transfer the learned category-level knowledge to the target domain, enabling the assignment of category labels to features. The memory bank is updated asynchronously with these assigned features to mitigate source domain bias. Extensive experiments conducted on diverse domain datasets demonstrate that the proposed method consistently outperforms state-of-the-art AOOD approaches, achieving performance gains of 1.1 to 5.5 mAP. Code is available at https://github.com/Jandsome/CCKM.

The goal of image cropping is to identify visually appealing crops in an image. Conventional methods are trained on specific datasets and fail to adapt to new requirements. Recent breakthroughs in large vision-language models (VLMs) enable visual in-context learning without explicit training. However, downstream tasks with VLMs remain under explored. In this paper, we propose an effective approach to leverage VLMs for image cropping. First, we propose an efficient prompt retrieval mechanism for image cropping to automate the selection of in-context examples. Second, we introduce an iterative refinement strategy to iteratively enhance the predicted crops. The proposed framework, we refer to as Cropper, is applicable to a wide range of cropping tasks, including free-form cropping, subject-aware cropping, and aspect ratio-aware cropping. Extensive experiments demonstrate that Cropper significantly outperforms state-of-the-art methods across several benchmarks.

Aligning text-to-image (T2I) diffusion models with preference optimization is valuable for human-annotated datasets, but the heavy cost of manual data collection limits scalability. Using reward models offers an alternative, however, current preference optimization methods fall short in exploiting the rich information, as they only consider pairwise preference distribution. Furthermore, they lack generalization to multi-preference scenarios and struggle to handle inconsistencies between rewards. To address this, we present Calibrated Preference Optimization (CaPO), a novel method to align T2I diffusion models by incorporating the general preference from multiple reward models without human annotated data. The core of our approach involves a reward calibration method to approximate the general preference by computing the expected win-rate against the samples generated by the pretrained models. Additionally, we propose a frontier-based pair selection method that effectively manages the multi-preference distribution by selecting pairs from Pareto frontiers. Finally, we use regression loss to fine-tune diffusion models to match the difference between calibrated rewards of a selected pair. Experimental results show that CaPO consistently outperforms prior methods, such as Direct Preference Optimization (DPO), in both single and multi-reward settings validated by evaluation on T2I benchmarks, including GenEval and T2I-Compbench.

Current object detectors often suffer performance degradation when applied to cross-domain scenarios, particularly under challenging visual conditions such as nighttime scenes. This is primarily due to the I3 problems: Inadequate sampling of instance-level features, Indistinguishable feature representation across domains and Inaccurate generation for identical category participation. To address these challenges, we propose a domain-adaptive detection framework that enables robust generalization across different visual domains without introducing any additional inference overhead. The framework comprises three key components. Specifically, the centerness-category consistency sampler alleviates inadequate sampling by selecting representative instance-level features, while the paired centerness consistency loss enforces alignment between classification and localization. Second, VLM-based orthogonality enhancement leverages frozen vision-language encoders with an orthogonal projection loss to improve cross-domain feature distinguishability. Third, hallucination feature generator synthesizes robust instance-level features for missing categories, ensuring balanced category participation across domains. Extensive experiments on multiple datasets covering various domain adaptation and generalization settings demonstrate that our method consistently outperforms state-of-the-art detectors, achieving up to 5.5 mAP improvement, with particularly strong gains in nighttime adaptation.

Inconsistent accuracy between classification and localization tasks is a common challenge in modern object detection. Task decoupling, which employs distinct features or labeling strategies for each task, is a widely used approach to address this issue. Although it has led to noteworthy advancements, this approach is insufficient as it neglects task interdependence and lacks an explicit consistency constraint. To bridge this gap, this paper proposes the Progressive Semi-Decoupled Detector (ProSDD) to enhance both classification and localization accuracy. Specifically, a new detection head is designed that incorporates feature suppression and enhancement mechanism (FSEM) and bidirectional interaction module (BIM). Compared with the decoupled head, it not only filters out task-irrelevant information and enhances task-related information, but also avoids excessive decoupling at the feature level. Moreover, both FSEM and BIM are used multiple times, thus forming a progressive semi-decoupled head. Then, a novel consistency loss is proposed and integrated into the loss function of object detection, ensuring harmonic performance in classification and localization. Experimental results demonstrate that the proposed ProSDD effectively alleviates inconsistent accuracy and achieves high-quality object detection. Taking the pretrained ResNet-50 as the backbone, ProSDD achieves a remarkable 43.3 AP on the MS COCO dataset, surpassing contemporary state-of-the-art detectors by a substantial margin under the equivalent configurations. Code is available athttps://github.com/HB-X/ProSDD.

Recent Text-to-Image (T2I) generation models such as Stable Diffusion and Imagen have made significant progress in generating high-resolution images based on text descriptions. However, many generated images still suffer from issues such as artifacts/implausibility, misalignment with text descriptions, and low aesthetic quality. Inspired by the success of Reinforcement Learning with Human Feedback (RLHF) for large language models, prior works collected human-provided scores as feedback on generated images and trained a reward model to improve the T2I generation. In this paper, we enrich the feedback signal by (i) marking image regions that are implausible or misaligned with the text, and (ii) annotating which words in the text prompt are misrepresented or missing on the image. We collect such rich human feedback on 18K generated images (RichHF-18K) and train a multimodal transformer to predict the rich feedback automatically. We show that the predicted rich human feedback can be leveraged to improve image generation, for example, by selecting high-quality training data to finetune and improve the generative models, or by creating masks with predicted heatmaps to inpaint the problematic regions. Notably, the improvements generalize to models (Muse) beyond those used to generate the images on which human feedback data were collected (Stable Diffusion variants). The RichHF-18K data set will be released in our GitHub repository: https://github.com/google-research/google-research/tree/master/richhf_18k.

Diffusion models generate new samples by progressively decreasing the noise from the initially provided random distribution. This inference procedure generally utilizes a trained neural network numerous times to obtain the final output, creating significant latency and energy consumption on digital electronic hardware such as GPUs. In this study, we demonstrate that the propagation of a light beam through a transparent medium can be programmed to implement a denoising diffusion model on image samples. This framework projects noisy image patterns through passive diffractive optical layers, which collectively only transmit the predicted noise term in the image. The optical transparent layers, which are trained with an online training approach, backpropagating the error to the analytical model of the system, are passive and kept the same across different steps of denoising. Hence this method enables high-speed image generation with minimal power consumption, benefiting from the bandwidth and energy efficiency of optical information processing.

Domain adaptive object detection (DAOD) aims to develop a detector trained on labeled source domains to identify objects in unlabeled target domains. A primary challenge in DAOD is the domain shift problem. Most existing methods learn domain-invariant features within single domain embedding space, often resulting in heavy model biases due to the intrinsic data properties of source domains. To mitigate the model biases, this paper proposes VLDadaptor, a domain adaptive object detector based on vision-language models (VLMs) distillation. Firstly, the proposed method integrates domain-mixed contrastive knowledge distillation between the visual encoder of CLIP and the detector by transferring category-level instance features, which guarantees the detector can extract domain-invariant visual instance features across domains. Then, VLDadaptor employs domain-mixed consistency distillation between the text encoder of CLIP and detector by aligning text prompt embeddings with visual instance features, which helps to maintain the category-level feature consistency among the detector, text encoder and the visual encoder of VLMs. Finally, the proposed method further promotes the adaptation ability by adopting a prompt-based memory bank to generate semantic-complete features for graph matching. These contributions enable VLDadaptor to extract visual features into the visual-language embedding space without any evident model bias towards specific domains. Extensive experimental results demonstrate that the proposed method achieves state-of-the-art performance on Pascal VOC to Clipart adaptation tasks and exhibits high accuracy on driving scenario tasks with significantly less training time.

Assessing the aesthetics of an image is challenging, as it is influenced by multiple factors including composition, color, style, and high-level semantics. Existing image aesthetic assessment (IAA) methods primarily rely on human-labeled rating scores, which oversimplify the visual aesthetic information that humans perceive. Conversely, user comments offer more comprehensive information and are a more natural way to express human opinions and preferences regarding image aesthetics. In light of this, we propose learning image aesthetics from user comments, and exploring vision-language pretraining methods to learn multimodal aesthetic representations. Specifically, we pretrain an image-text encoder-decoder model with image-comment pairs, using contrastive and generative objectives to learn rich and generic aesthetic semantics without human labels. To efficiently adapt the pretrained model for downstream IAA tasks, we further propose a lightweight rank-based adapter that employs text as an anchor to learn the aesthetic ranking concept. Our results show that our pretrained aesthetic vision-language model outperforms prior works on image aesthetic captioning over the AVA-Captions dataset, and it has powerful zero-shot capability for aesthetic tasks such as zero-shot style classification and zero-shot IAA, surpassing many supervised baselines. With only minimal finetuning parameters using the proposed adapter module, our model achieves state-of-the-art IAA performance over the AVA dataset.11Our model is available at https://github.com/google-research/google-research/tree/master/VILA

This paper presents a framework to identify whether a document image is perturbed with glare. Glare identification for document images is particularly challenging because of predominantly white background and dearth of training dataset. We addresses the dataset bottleneck by introducing a glare synthesis framework to generate a large training dataset. The proposed training model consists of a global deep neural network supplemented by extracted localized feature. To our best knowledge, this is one of the first works towards classifying document image for presence of glare. Experiments on real glare dataset showcase benefits of combined global and local features and also outperform recent glare segmentation model adapted for the classification task.

UGC video quality assessment (UGC-VQA) is a challenging research topic due to the high video diversity and limited public UGC quality datasets. State-of-the-art (SOTA) UGC quality models tend to use high complexity models, and rarely discuss the trade-off among complexity, accuracy, and generalizability. We propose a new perspective on UGC-VQA, and show that model complexity may not be critical to the performance, whereas a more diverse dataset is essential to train a better model. We illustrate this by using a light weight model, UVQ-lite, which has higher efficiency and better generalizability (less overfitting) than baseline SOTA models. We also propose a new way to analyze the sufficiency of the training set, by leveraging UVQ’s comprehensive features. Our results motivate a new perspective about the future of UGC-VQA research, which we believe is headed toward more efficient models and more diverse datasets.

Single domain generalization aims to learn a model that performs well on many unseen domains with only one domain data for training. Existing works focus on studying the adversarial domain augmentation (ADA) to improve the model’s generalization capability. The impact on domain generalization of the statistics of normalization layers is still underinvestigated. In this paper, we propose a generic normalization approach, adaptive standardization and rescaling normalization (ASR-Norm), to complement the missing part in previous works. ASR-Norm learns both the standardization and rescaling statistics via neural networks. This new form of normalization can be viewed as a generic form of the traditional normalizations. When trained with ADA, the statistics in ASR-Norm are learned to be adaptive to the data coming from different domains, and hence improves the model generalization performance across domains, especially on the target domain with large discrepancy from the source domain. The experimental results show that ASR-Norm can bring consistent improvement to the state-of-the-art ADA approaches by 1.6%, 2.7%, and 6.3% averagely on the Digits, CIFAR-10-C, and PACS benchmarks, respectively. As a generic tool, the improvement introduced by ASR-Norm is agnostic to the choice of ADA methods.

Video quality assessment for User Generated Content (UGC) is an important topic in both industry and academia. Most existing methods only focus on one aspect of the perceptual quality assessment, such as technical quality or compression artifacts. In this paper, we create a large scale dataset to comprehensively investigate characteristics of generic UGC video quality. Besides the subjective ratings and content labels of the dataset, we also propose a DNN-based framework to thoroughly analyze importance of content, technical quality, and compression level in perceptual quality. Our model is able to provide quality scores as well as human-friendly quality indicators, to bridge the gap between low level video signals to human perceptual quality. Experimental results show that our model achieves state-of-the-art correlation with Mean Opinion Scores (MOS).

Image quality assessment (IQA) is an important research topic for understanding and improving visual experience. The current state-of-the-art IQA methods are based on convolutional neural networks (CNNs). The performance of CNN-based models is often compromised by the fixed shape constraint in batch training. To accommodate this, the input images are usually resized and cropped to a fixed shape, causing image quality degradation. To address this, we design a multi-scale image quality Transformer (MUSIQ) to process native resolution images with varying sizes and aspect ratios. With a multi-scale image representation, our proposed method can capture image quality at different granularities. Furthermore, a novel hash-based 2D spatial embedding and a scale embedding is proposed to support the positional embedding in the multi-scale representation. Experimental results verify that our method can achieve state-of-the-art performance on multiple large scale IQA datasets such as PaQ-2-PiQ [41], SPAQ [11], and KonIQ-10k [16].1

Learning multiple domains/tasks with a single model is important for improving data efficiency and lowering inference cost for numerous vision tasks, especially on resource-constrained mobile devices. However, hand-crafting a multi-domain/task model can be both tedious and challenging. This paper proposes a novel approach to automatically learn a multi-path network for multi-domain visual classification on mobile devices. The proposed multi-path network is learned from neural architecture search by applying one reinforcement learning controller for each domain to select the best path in the super-network created from a MobileNetV3-like search space. An adaptive balanced domain prioritization algorithm is proposed to balance optimizing the joint model on multiple domains simultaneously. The determined multi-path model selectively shares parameters across domains in shared nodes while keeping domain-specific parameters within non-shared nodes in individual domain paths. This approach effectively reduces the total number of parameters and FLOPS, encouraging positive knowledge transfer while mitigating negative interference across domains. Extensive evaluations on the Visual Decathlon dataset demonstrate that the proposed multi-path model achieves state-of-the-art performance in terms of accuracy, model size, and FLOPS against other approaches using MobileNetV3-like architectures. Furthermore, the proposed method improves average accuracy over learning single-domain models individually, and reduces the total number of parameters and FLOPS by 78% and 32% respectively, compared to the approach that simply bundles single-domain models for multi-domain learning.