Junxing Ren

Chenghua Gao, Min Li, Jianshuo Liu et al.

Video Moment Retrieval (VMR) aims to retrieve relevant moments of an untrimmed video corresponding to the query. While cross-modal interaction approaches have shown progress in filtering out query-irrelevant information in videos, they assume the precise alignment between the query semantics and the corresponding video moments, potentially overlooking the misunderstanding of the natural language semantics. To address this challenge, we propose a novel model called \textit{QD-VMR}, a query debiasing model with enhanced contextual understanding. Firstly, we leverage a Global Partial Aligner module via video clip and query features alignment and video-query contrastive learning to enhance the cross-modal understanding capabilities of the model. Subsequently, we employ a Query Debiasing Module to obtain debiased query features efficiently, and a Visual Enhancement module to refine the video features related to the query. Finally, we adopt the DETR structure to predict the possible target video moments. Through extensive evaluations of three benchmark datasets, QD-VMR achieves state-of-the-art performance, proving its potential to improve the accuracy of VMR. Further analytical experiments demonstrate the effectiveness of our proposed module. Our code will be released to facilitate future research.

Ruwen Bai, Min Li, Bo Meng et al.

Graph convolutional networks (GCNs) have emerged as dominant methods for skeleton-based action recognition. However, they still suffer from two problems, namely, neighborhood constraints and entangled spatiotemporal feature representations. Most studies have focused on improving the design of graph topology to solve the first problem but they have yet to fully explore the latter. In this work, we design a disentangled spatiotemporal transformer (DSTT) block to overcome the above limitations of GCNs in three steps: (i) feature disentanglement for spatiotemporal decomposition;(ii) global spatiotemporal attention for capturing correlations in the global context; and (iii) local information enhancement for utilizing more local information. Thereon, we propose a novel architecture, named Hierarchical Graph Convolutional skeleton Transformer (HGCT), to employ the complementary advantages of GCN (i.e., local topology, temporal dynamics and hierarchy) and Transformer (i.e., global context and dynamic attention). HGCT is lightweight and computationally efficient. Quantitative analysis demonstrates the superiority and good interpretability of HGCT.

Yang Yang, Min Li, Bo Meng et al.

Object detection aims to localize and classify the objects in a given image, and these two tasks are sensitive to different object regions. Therefore, some locations predict high-quality bounding boxes but low classification scores, and some locations are quite the opposite. A misalignment exists between the two tasks, and their features are spatially entangled. In order to solve the misalignment problem, we propose a plug-in Spatial-disentangled and Task-aligned operator (SALT). By predicting two task-aware point sets that are located in each task's sensitive regions, SALT can reassign features from those regions and align them to the corresponding anchor point. Therefore, features for the two tasks are spatially aligned and disentangled. To minimize the difference between the two regression stages, we propose a Self-distillation regression (SDR) loss that can transfer knowledge from the refined regression results to the coarse regression results. On the basis of SALT and SDR loss, we propose SALT-Net, which explicitly exploits task-aligned point-set features for accurate detection results. Extensive experiments on the MS-COCO dataset show that our proposed methods can consistently boost different state-of-the-art dense detectors by $\sim$2 AP. Notably, SALT-Net with Res2Net-101-DCN backbone achieves 53.8 AP on the MS-COCO test-dev.

Yang Yang, Min Li, Bo Meng et al.

Object detection can be regarded as a pixel clustering task, and its boundary is determined by four extreme points (leftmost, top, rightmost, and bottom). However, most studies focus on the center or corner points of the object, which are actually conditional results of the extreme points. In this paper, we present an Extreme-Point-Prediction- Based object detector (EPP-Net), which directly regresses the relative displacement vector between each pixel and the four extreme points. We also propose a new metric to measure the similarity between two groups of extreme points, namely, Extreme Intersection over Union (EIoU), and incorporate this EIoU as a new regression loss. Moreover, we propose a novel branch to predict the EIoU between the ground-truth and the prediction results, and take it as the localization confidence to filter out poor detection results. On the MS-COCO dataset, our method achieves an average precision (AP) of 44.0% with ResNet-50 and an AP of 50.3% with ResNeXt-101-DCN. The proposed EPP-Net provides a new method to detect objects and outperforms state-of-the-art anchor-free detectors.