PCA based appearance model for tracking

In this project, we studied PCA based appearance model in visual tracking. We implemented PCA AM module in MTF framework and attempted different ways of formulating PCA. In the experiment sections, we demonstrate the tracking performance of offline PCA when combined with different search methods. PCA + Forward Compositional has shown great tracking accuracy and robustness. It indicates a good potential for online PCA method. In the end, we suggests the future work in how we can use it with nearest neighbor to improve the accuracy.