Abstract
As a classic framework for visual object tracking, the Siamese convolutional neural network has received widespread attention from the research community. This method uses a convolutional neural network to obtain the object features and to match them with the search area features to achieve object tracking. In this work, we observe that the contribution of each convolution kernel in the convolutional neural network for object tracking tasks is different. We propose an object-aware convolution kernel attention mechanism. Based on the characteristics of each object, the convolution kernel features are dynamically weighted to improve the expression ability of object features. The experiments performed using OTB and VOT benchmark datasets show that the performance of the tracking method fused with the convolution kernel attention mechanism is significantly better compared with the original method. Moreover, the attention mechanism can also be integrated with other tracking frameworks as an independent module to improve the performance.
Highlights
Visual object tracking is a fundamental task in the field of computer vision
The object tracking methods based on Siamese convolutional neural networks have been developed rapidly due to their simple structures and efficiency
We present the convolution kerSiamese network uses a series of image pairs to train the network, the attention module nel-oriented “Squeeze-and-Excitation” attention mechanism [19] ability tobased simultaneously learnnetwork the weight of each convolution kernel in the tention on the Siamese object tracking framework
Summary
Visual object tracking is a fundamental task in the field of computer vision. It has a wide range of application prospects in autopilot, robot navigation, drone cruise, intelligent security, human–computer interaction, etc. Based on the aforementioned analysis, in this work, we propose an adaptive convolution kernel attention mechanism for visual object tracking tasks. The paper analyzes the problems with the offline training-based attention mechanism in the tracking task and finds out the reason why this attention method is not suitable for object tracking. The proposed module improves the specificity of object features for different individuals based on the difference between each convolution kernel feature. The proposed attention mechanism can be used as a general convolution kernel attention module and has good versatility It can be used as a “plug and play” module by embedding it in various object tracking methods based on convolution feature matching, significantly improving the tracking performance of the original method. We combine the proposed convolution kernel attention module with classic object tracking methods. We perform experiments using OTB and VOT benchmark datasets to verify that the proposed technique significantly improves the object tracking performance
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
