TransVOD: End-to-End Video Object Detection with Spatial-Temporal Transformers

TL;DR

TransVOD introduces an end-to-end video object detection system based on spatial-temporal Transformers, eliminating many traditional components and achieving state-of-the-art accuracy and speed on the ImageNet VID dataset.

Contribution

This paper presents the first end-to-end VOD system using spatial-temporal Transformers, removing hand-crafted components and introducing improved models TransVOD++, and TransVOD Lite.

Findings

Boosts deformable DETR baseline by 3-4% mAP on ImageNet VID

TransVOD++ achieves 90.0% mAP, setting new state-of-the-art

TransVOD Lite balances speed (30 FPS) and accuracy (83.7% mAP)

Abstract

Detection Transformer (DETR) and Deformable DETR have been proposed to eliminate the need for many hand-designed components in object detection while demonstrating good performance as previous complex hand-crafted detectors. However, their performance on Video Object Detection (VOD) has not been well explored. In this paper, we present TransVOD, the first end-to-end video object detection system based on spatial-temporal Transformer architectures. The first goal of this paper is to streamline the pipeline of VOD, effectively removing the need for many hand-crafted components for feature aggregation, e.g., optical flow model, relation networks. Besides, benefited from the object query design in DETR, our method does not need complicated post-processing methods such as Seq-NMS. In particular, we present a temporal Transformer to aggregate both the spatial object queries and the feature memories of each frame. Our temporal transformer consists of two components: Temporal Query Encoder (TQE) to fuse object queries, and Temporal Deformable Transformer Decoder (TDTD) to obtain current frame detection results. These designs boost the strong baseline deformable DETR by a significant margin (3%-4% mAP) on the ImageNet VID dataset. Then, we present two improved versions of TransVOD including TransVOD++ and TransVOD Lite. The former fuses object-level information into object query via dynamic convolution while the latter models the entire video clips as the output to speed up the inference time. We give detailed analysis of all three models in the experiment part. In particular, our proposed TransVOD++ sets a new state-of-the-art record in terms of accuracy on ImageNet VID with 90.0% mAP. Our proposed TransVOD Lite also achieves the best speed and accuracy trade-off with 83.7% mAP while running at around 30 FPS on a single V100 GPU device.