MixFormer: End-to-End Tracking with Iterative Mixed Attention

Abstract

Visual object tracking often employs a multi-stage pipeline of feature extraction, target information integration, and bounding box estimation. To simplify this pipeline and unify the process of feature extraction and target information integration, in this paper, we present a compact tracking framework, termed as MixFormer, built upon transformers. Our core design is to utilize the flexibility of attention operations, and propose a Mixed Attention Module (MAM) for simultaneous feature extraction and target information integration. This synchronous modeling scheme allows to extract target-specific discriminative features and perform extensive communication between target and search area. Based on MAM, we build our MixFormer trackers simply by stacking multiple MAMs and placing a localization head on top. Specifically, we instantiate two types of MixFormer trackers, a hierarchical tracker MixCvT, and a non-hierarchical tracker MixViT. For these two trackers, we investigate a series of pre-training methods and uncover the different behaviors between supervised pre-training and self-supervised pre-training in our MixFormer trackers. We also extend the masked pre-training to our MixFormer trackers and design the competitive TrackMAE pre-training technique. Finally, to handle multiple target templates during online tracking, we devise an asymmetric attention scheme in MAM to reduce computational cost, and propose an effective score prediction module to select high-quality templates. Our MixFormer trackers set a new state-of-the-art performance on seven tracking benchmarks, including LaSOT, TrackingNet, VOT2020, GOT-10k, OTB100 and UAV123. In particular, our MixViT-L achieves AUC score of 73.3% on LaSOT, 86.1% on TrackingNet, EAO of 0.584 on VOT2020, and AO of 75.7% on GOT-10k. Code and trained models are publicly available at https://github.com/MCG-NJU/MixFormer.

Enhancing Single Object Tracking with Temporal Information

“…Trackers employing transformers can be categorized into two classes: CNN transformer-based trackers like STARK [6], AiATrack [25], TransT [18] and TrTr [26]. And fully-transformer based trackers like SwinTrack [27], MixFormer [19], Pro-ContEXT [28] and VideoTrack [29]. The state-ofthe-art tracker tends to utilize a fully-transformer architecture as it achieves the best performance [30], while CNN transformer architecture offers a balanced performance between speed and accuracy [30].…”

“…Additionally, transformers [15], originally applied to natural language processing (NLP), which have showed excellent performance in various computer vision tasks like ViT [16] and DETR [17], have gained attention in the tracking field. Some transformer-based trackers utilize CNNs as a backbone [18] [6], while others solely rely on attention mechanisms [19] [20] [7]. However, most existing trackers primarily rely on spatial information extracted from the object appearance, neglecting valuable temporal information such as motion model and changes in object appearance.…”