The classification and regression head are both indispensable components to build up a dense object detector, which are usually supervised by the same training samples and thus expected to have consistency with each other for detecting objects accura
tely in the detection pipeline. In this paper, we break the convention of the same training samples for these two heads in dense detectors and explore a novel supervisory paradigm, termed as Mutual Supervision (MuSu), to respectively and mutually assign training samples for the classification and regression head to ensure this consistency. MuSu defines training samples for the regression head mainly based on classification predicting scores and in turn, defines samples for the classification head based on localization scores from the regression head. Experimental results show that the convergence of detectors trained by this mutual supervision is guaranteed and the effectiveness of the proposed method is verified on the challenging MS COCO benchmark. We also find that tiling more anchors at the same location benefits detectors and leads to further improvements under this training scheme. We hope this work can inspire further researches on the interaction of the classification and regression task in detection and the supervision paradigm for detectors, especially separately for these two heads.
Temporal grounding aims to temporally localize a video moment in the video whose semantics are related to a given natural language query. Existing methods typically apply a detection or regression pipeline on the fused representation with a focus on
designing complicated heads and fusion strategies. Instead, from a perspective on temporal grounding as a metric-learning problem, we present a Dual Matching Network (DMN), to directly model the relations between language queries and video moments in a joint embedding space. This new metric-learning framework enables fully exploiting negative samples from two new aspects: constructing negative cross-modal pairs from a dual matching scheme and mining negative pairs across different videos. These new negative samples could enhance the joint representation learning of two modalities via cross-modal pair discrimination to maximize their mutual information. Experiments show that DMN achieves highly competitive performance compared with state-of-the-art methods on four video grounding benchmarks. Based on DMN, we present a winner solution for STVG challenge of the 3rd PIC workshop. This suggests that metric-learning is still a promising method for temporal grounding via capturing the essential cross-modal correlation in a joint embedding space.
Deep learning has achieved remarkable progress for visual recognition on large-scale balanced datasets but still performs poorly on real-world long-tailed data. Previous methods often adopt class re-balanced training strategies to effectively allevia
te the imbalance issue, but might be a risk of over-fitting tail classes. The recent decoupling method overcomes over-fitting issues by using a multi-stage training scheme, yet, it is still incapable of capturing tail class information in the feature learning stage. In this paper, we show that soft label can serve as a powerful solution to incorporate label correlation into a multi-stage training scheme for long-tailed recognition. The intrinsic relation between classes embodied by soft labels turns out to be helpful for long-tailed recognition by transferring knowledge from head to tail classes. Specifically, we propose a conceptually simple yet particularly effective multi-stage training scheme, termed as Self Supervised to Distillation (SSD). This scheme is composed of two parts. First, we introduce a self-distillation framework for long-tailed recognition, which can mine the label relation automatically. Second, we present a new distillation label generation module guided by self-supervision. The distilled labels integrate information from both label and data domains that can model long-tailed distribution effectively. We conduct extensive experiments and our method achieves the state-of-the-art results on three long-tailed recognition benchmarks: ImageNet-LT, CIFAR100-LT and iNaturalist 2018. Our SSD outperforms the strong LWS baseline by from $2.7%$ to $4.5%$ on various datasets. The code is available at https://github.com/MCG-NJU/SSD-LT.
This paper deals with a challenging task of video scene graph generation (VidSGG), which could serve as a structured video representation for high-level understanding tasks. We present a new {em detect-to-track} paradigm for this task by decoupling t
he context modeling for relation prediction from the complicated low-level entity tracking. Specifically, we design an efficient method for frame-level VidSGG, termed as {em Target Adaptive Context Aggregation Network} (TRACE), with a focus on capturing spatio-temporal context information for relation recognition. Our TRACE framework streamlines the VidSGG pipeline with a modular design, and presents two unique blocks of Hierarchical Relation Tree (HRTree) construction and Target-adaptive Context Aggregation. More specific, our HRTree first provides an adpative structure for organizing possible relation candidates efficiently, and guides context aggregation module to effectively capture spatio-temporal structure information. Then, we obtain a contextualized feature representation for each relation candidate and build a classification head to recognize its relation category. Finally, we provide a simple temporal association strategy to track TRACE detected results to yield the video-level VidSGG. We perform experiments on two VidSGG benchmarks: ImageNet-VidVRD and Action Genome, and the results demonstrate that our TRACE achieves the state-of-the-art performance. The code and models are made available at url{https://github.com/MCG-NJU/TRACE}.
Scene graph generation (SGG) is to detect entity pairs with their relations in an image. Existing SGG approaches often use multi-stage pipelines to decompose this task into object detection, relation graph construction, and dense or dense-to-sparse r
elation prediction. Instead, from a perspective on SGG as a direct set prediction, this paper presents a simple, sparse, and unified framework for relation detection, termed as Structured Sparse R-CNN. The key to our method is a set of learnable triplet queries and structured triplet detectors which could be jointly optimized from the training set in an end-to-end manner. Specifically, the triplet queries encode the general prior for entity pair locations, categories, and their relations, and provide an initial guess of relation detection for subsequent refinement. The triplet detector presents a cascaded dynamic head design to progressively refine the results of relation detection. In addition, to relieve the training difficulty of Structured Sparse R-CNN, we propose a relaxed and enhanced training strategy based on knowledge distillation from a Siamese Sparse R-CNN. We also propose adaptive focusing parameter and average logit approach for imbalance data distribution. We perform experiments on two benchmarks: Visual Genome and Open Images, and the results demonstrate that our method achieves the state-of-the-art performance. Meanwhile, we perform in-depth ablation studies to provide insights on our structured modeling in triplet detector design and training strategies.
The ultrasound (US) screening of the infant hip is vital for the early diagnosis of developmental dysplasia of the hip (DDH). The US diagnosis of DDH refers to measuring alpha and beta angles that quantify hip joint development. These two angles are
calculated from key anatomical landmarks and structures of the hip. However, this measurement process is not trivial for sonographers and usually requires a thorough understanding of complex anatomical structures. In this study, we propose a multi-task framework to learn the relationships among landmarks and structures jointly and automatically evaluate DDH. Our multi-task networks are equipped with three novel modules. Firstly, we adopt Mask R-CNN as the basic framework to detect and segment key anatomical structures and add one landmark detection branch to form a new multi-task framework. Secondly, we propose a novel shape similarity loss to refine the incomplete anatomical structure prediction robustly and accurately. Thirdly, we further incorporate the landmark-structure consistent prior to ensure the consistency of the bony rim estimated from the segmented structure and the detected landmark. In our experiments, 1,231 US images of the infant hip from 632 patients are collected, of which 247 images from 126 patients are tested. The average errors in alpha and beta angles are 2.221 degrees and 2.899 degrees. About 93% and 85% estimates of alpha and beta angles have errors less than 5 degrees, respectively. Experimental results demonstrate that the proposed method can accurately and robustly realize the automatic evaluation of DDH, showing great potential for clinical application.
Temporal action localization (TAL) is an important and challenging problem in video understanding. However, most existing TAL benchmarks are built upon the coarse granularity of action classes, which exhibits two major limitations in this task. First
, coarse-level actions can make the localization models overfit in high-level context information, and ignore the atomic action details in the video. Second, the coarse action classes often lead to the ambiguous annotations of temporal boundaries, which are inappropriate for temporal action localization. To tackle these problems, we develop a novel large-scale and fine-grained video dataset, coined as FineAction, for temporal action localization. In total, FineAction contains 103K temporal instances of 106 action categories, annotated in 17K untrimmed videos. FineAction introduces new opportunities and challenges for temporal action localization, thanks to its distinct characteristics of fine action classes with rich diversity, dense annotations of multiple instances, and co-occurring actions of different classes. To benchmark FineAction, we systematically investigate the performance of several popular temporal localization methods on it, and deeply analyze the influence of short-duration and fine-grained instances in temporal action localization. We believe that FineAction can advance research of temporal action localization and beyond.
Frame sampling is a fundamental problem in video action recognition due to the essential redundancy in time and limited computation resources. The existing sampling strategy often employs a fixed frame selection and lacks the flexibility to deal with
complex variations in videos. In this paper, we present a simple, sparse, and explainable frame sampler, termed as Motion-Guided Sampler (MGSampler). Our basic motivation is that motion is an important and universal signal that can drive us to adaptively select frames from videos. Accordingly, we propose two important properties in our MGSampler design: motion sensitive and motion uniform. First, we present two different motion representations to enable us to efficiently distinguish the motion-salient frames from the background. Then, we devise a motion-uniform sampling strategy based on the cumulative motion distribution to ensure the sampled frames evenly cover all the important segments with high motion salience. Our MGSampler yields a new principled and holistic sampling scheme, that could be incorporated into any existing video architecture. Experiments on five benchmarks demonstrate the effectiveness of our MGSampler over the previous fixed sampling strategies, and its generalization power across different backbones, video models, and datasets.
Accurate tracking is still a challenging task due to appearance variations, pose and view changes, and geometric deformations of target in videos. Recent anchor-free trackers provide an efficient regression mechanism but fail to produce precise bound
ing box estimation. To address these issues, this paper repurposes a Transformer-alike regression branch, termed as Target Transformed Regression (TREG), for accurate anchor-free tracking. The core to our TREG is to model pair-wise relation between elements in target template and search region, and use the resulted target enhanced visual representation for accurate bounding box regression. This target contextualized representation is able to enhance the target relevant information to help precisely locate the box boundaries, and deal with the object deformation to some extent due to its local and dense matching mechanism. In addition, we devise a simple online template update mechanism to select reliable templates, increasing the robustness for appearance variations and geometric deformations of target in time. Experimental results on visual tracking benchmarks including VOT2018, VOT2019, OTB100, GOT10k, NFS, UAV123, LaSOT and TrackingNet demonstrate that TREG obtains the state-of-the-art performance, achieving a success rate of 0.640 on LaSOT, while running at around 30 FPS. The code and models will be made available at https://github.com/MCG-NJU/TREG.
Temporal action proposal generation is an important and challenging task in video understanding, which aims at detecting all temporal segments containing action instances of interest. The existing proposal generation approaches are generally based on
pre-defined anchor windows or heuristic bottom-up boundary matching strategies. This paper presents a simple and efficient framework (RTD-Net) for direct action proposal generation, by re-purposing a Transformer-alike architecture. To tackle the essential visual difference between time and space, we make three important improvements over the original transformer detection framework (DETR). First, to deal with slowness prior in videos, we replace the original Transformer encoder with a boundary attentive module to better capture long-range temporal information. Second, due to the ambiguous temporal boundary and relatively sparse annotations, we present a relaxed matching scheme to relieve the strict criteria of single assignment to each groundtruth. Finally, we devise a three-branch head to further improve the proposal confidence estimation by explicitly predicting its completeness. Extensive experiments on THUMOS14 and ActivityNet-1.3 benchmarks demonstrate the effectiveness of RTD-Net, on both tasks of temporal action proposal generation and temporal action detection. Moreover, due to its simplicity in design, our framework is more efficient than previous proposal generation methods, without non-maximum suppression post-processing. The code and models are made available at https://github.com/MCG-NJU/RTD-Action.
