Multiscale Convolutional Neural Networks for Hand Detection

Unconstrained hand detection in still images plays an important role in many hand-related vision problems, for example, hand tracking, gesture analysis, human action recognition and human-machine interaction, and sign language recognition. Although hand detection has been extensively studied for decades, it isstill a challenging task with many problems to be tackled. Thec ontributing factors for this complexity include heavy occlusion, low resolution, varying illumination conditions, different hand gestures, and the complex interactions between hands and objects or other hands. In this paper, we propose a multi scale deepl earning model for unconstrained hand detection in still images. Deep learning models, and deep convolutional neural networks(CNNs) in particular, have achieved state-of-the-art performances in many vision benchmarks. Developed from the region-basedCNN (R-CNN) model, we propose a hand detection scheme based on candidate regions generated by a generic region proposalalgorithm, followed by multiscale information fusion from the popular VGG16 model. Two benchmark datasets were applied tovalidate the proposed method, namely, the Oxford Hand Detection Data set and the VIVA Hand Detection Challenge. We achieve dstate-of-the-art results on the Oxford Hand Detection Data set and had satisfactory performance in the VIVA Hand Detection Challenge.