A rough to fine hand detection method using deep neural networks

摘要

A detection process is provided for recognizing one or more regions of an image that include one or more hands of one or more objects. The detection process uses a rough positioning network to roughly position one or more segments in the image that include the part (s) of the object (s) of interest in the image. You can start with The detection process can then combine these segments to obtain one or more regions that capture the hand (s) of interest in the image (s). The combined region (s) is then fed to a mesh-based deep neural network to capture only the captured hand (s) of the captured object (s) in the image. The containing region (s) can be finely detected.