Abstract:High computational power and significant time are usually needed to train a deep learning based tracker on large datasets. Depending on many factors, training might not always be an option. In this paper, we propose a framework with two ideas on Siamese-based trackers. (i) Extending number of templates in a way that removes the need to retrain the network and (ii) a lightweight temporal network with a novel architecture focusing on both local and global information that can be used independently from trackers. Most Siamese-based trackers only rely on the first frame as the ground truth for objects and struggle when the target's appearance changes significantly in subsequent frames in presence of similar distractors. Some trackers use multiple templates which mostly rely on constant thresholds to update, or they replace those templates that have low similarity scores only with more similar ones. Unlike previous works, we use adaptive thresholds that update the bag with similar templates as well as those templates which are slightly diverse. Adaptive thresholds also cause an overall improvement over constant ones. In addition, mixing feature maps obtained by each template in the last stage of networks removes the need to retrain trackers. Our proposed lightweight temporal network, CombiNet, learns the path history of different objects using only object coordinates and predicts target's potential location in the next frame. It is tracker independent and applying it on new trackers does not need further training. By implementing these ideas, trackers' performance improved on all datasets tested on, including LaSOT, LaSOT extension, TrackingNet, OTB100, OTB50, UAV123 and UAV20L. Experiments indicate the proposed framework works well with both convolutional and transformer-based trackers. The official python code for this paper will be publicly available upon publication.

Siamese Tracking With Adaptive Template-Updating Strategy

Siamese Tracking from Single Point Initialization

Occlusion-Aware Visual Object Tracking Based on Multi-template Updating Siamese Network

Dynamic template updating Siamese network based on status feedback with quality evaluation for visual object tracking

ASiam: Adaptive Siamese Regression Tracking with Adversarial Template Generation and Motion-Based Failure Recovery.

Distractor-Aware Visual Tracking by Online Siamese Network

Dual-template Siamese Network with Cross-Correlation Fusion for Tracking

Siamese Tracking Network with Spatial-Semantic-Aware Attention and Flexible Spatiotemporal Constraint

Adaptive distractor-aware for siamese tracking via enhancement confidence evaluator

Improving Siamese Based Trackers with Light or No Training through Multiple Templates and Temporal Network

SiamRDT: An Object Tracking Algorithm Based on a Reliable Dynamic Template

Learning Spatio-Temporal Convolutional Network for Real-Time Object Tracking

Siamese Tracking Network with Multi-attention Mechanism

AtptTrack: Asymmetric Transformer Tracker With Prior Templates

Template Pool Updating and Feature Combination Object Tracking

SiamDMU: Siamese Dual Mask Update Network for Visual Object Tracking

Visual Object Tracking with Adaptive Template Update and Global Search Augmentation

Masked and Dynamic Siamese Network for Robust Visual Tracking.

Adaptive Hierarchical Siamese Network For Object Tracking

GradNet: Gradient-Guided Network for Visual Object Tracking

Siamese Block Attention Network for Online Update Object Tracking