🤖 AI Summary
In video object detection, unimodal models—whether CNN- or ViT-based—struggle to jointly capture fine-grained local details and holistic spatiotemporal context. To address this, we propose a dual-path collaborative architecture: one path employs a spatiotemporal Transformer to model long-range temporal dependencies; the other adopts a graph-enhanced GraphFormer that explicitly encodes intra-frame object relationships and inter-frame correspondences. We further introduce a novel adaptive global-local feature fusion module that dynamically couples the complementary strengths of these two paradigms. The entire framework operates end-to-end on video frame sequences. Evaluated on ImageNet VID, our method achieves 86.5% mAP while maintaining real-time inference at 41.0 FPS on a single A100 GPU—setting a new state-of-the-art performance.
📝 Abstract
Video object detection has made significant progress in recent years thanks to convolutional neural networks (CNNs) and vision transformers (ViTs). Typically, CNNs excel at capturing local features but struggle to model global representations. Conversely, ViTs are adept at capturing long-range global features but face challenges in representing local feature details. Off-the-shelf video object detection methods solely rely on CNNs or ViTs to conduct feature aggregation, which hampers their capability to simultaneously leverage global and local information, thereby resulting in limited detection performance. In this paper, we propose a Transformer-GraphFormer Blender Network (TGBFormer) for video object detection, with three key technical improvements to fully exploit the advantages of transformers and graph convolutional networks while compensating for their limitations. First, we develop a spatial-temporal transformer module to aggregate global contextual information, constituting global representations with long-range feature dependencies. Second, we introduce a spatial-temporal GraphFormer module that utilizes local spatial and temporal relationships to aggregate features, generating new local representations that are complementary to the transformer outputs. Third, we design a global-local feature blender module to adaptively couple transformer-based global representations and GraphFormer-based local representations. Extensive experiments demonstrate that our TGBFormer establishes new state-of-the-art results on the ImageNet VID dataset. Particularly, our TGBFormer achieves 86.5% mAP while running at around 41.0 FPS on a single Tesla A100 GPU.