We present HARM, a three stage human action recognition algorithm for static images. The proposed methodology incorporates pose estimation, pose classification and scene classification to develop a complete understanding of humans in images with respect to their environments. The module utilizes a blend of modern state-of-the-art pose estimation and image classification algorithms, thereby resulting in accurate classifications of human poses and actions in static images.
CITATION STYLE
Fernandes, B., Alag, G., & Kumaar, S. (2019). HARM - The Human Action Recognition Module. In Lecture Notes in Computational Vision and Biomechanics (Vol. 34, pp. 419–433). Springer Netherlands. https://doi.org/10.1007/978-3-030-32040-9_43
Mendeley helps you to discover research relevant for your work.