Hand detection has many important applications in HCI, yet it is a challenging problem because the appearance of hands can vary greatly in images. In this paper, we propose a novel method for efficient pixel-level hand detection. Unlike previous method which assigns a binary label to every pixel independently, our method estimates a probability shape mask for a pixel using structured forests. This approach can better exploit hand shape information in the training data, and enforce shape constraints in the estimation. Aggregation of multiple predictions generated from neighboring pixels further improves the robustness of our method. We evaluate our method on both ego-centric videos and unconstrained still images. Experiment results show that our method can detect hands efficiently and outperform other state-of-the-art methods.
CITATION STYLE
Zhu, X., Jia, X., & Wong, K. Y. K. (2015). Pixel-level hand detection with shape-aware structured forests. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9006, pp. 64–78). Springer Verlag. https://doi.org/10.1007/978-3-319-16817-3_5
Mendeley helps you to discover research relevant for your work.