Terminus Group’s Adaptive Siamese Tracking with CLNet Enhances Motion Capture

Terminus Group’s adaptive Siamese tracking technology with Compact Latent Network (CLNet) has been proofed to empower tracking algorithms more capabilities of capturing motion moments, after an effectiveness and generalization ability testament.

This approach is now in an attempt to actual scene use.

What’s the issue?

Accurate tracking in real-world motion scenarios is often disrupted by factors such as similar-looking objects, lighting, occlusions, complex backgrounds, and changes in target size.

After an in-depth analysis for visual simulations and real tracking examples, it turns out that the failure cases in some challenging situations can be regarded as the issue of missing decisive samples in offline training. Since the samples in the initial (first) frame contain rich sequence-specific information, they can be considered as decisive samples to represent the whole sequence. 

To quickly adapt the base model to new scenes, a compact latent network is presented via fully using these decisive samples.

After the improvement, the approach enhances the robustness of tracking algorithms in dealing with complex scenarios include not only sport events but also autonomous driving, robot navigation, industrial automation, aerospace and more.

For original paper info, please refer to:

Adaptive Siamese Tracking With a Compact Latent Network | IEEE Journals & Magazine | IEEE Xplore