Though recently I have not much time for this repository maintainance possess, But fortunately (maybe you could get more details of this project) I wrote a blog (in Chinese) about the handpose investigation (traditional style and DL style) completed during this project and intermittently answered relevant questions raised by people who were interested in the comments section of the blog.

Handpose-WonderSeen-Net

1. DATABASE: RHD_published_v2.
2. DATABASE INFO: RGBD, four channels, pix-level label.
3. CODE: To be updated in 2 months.
4. ADDITION: This script isn't the final version of my work of gesture recognization. For that Code Management requires a lot of work, the script 'Wonderseen_HandPose_cnn_depth.py' is provided merely as reference to show the main thoughts of the process.

Network-Result

Reference

[1] Hand Gesture Recognition Based on Shape Parameters.

[2] Densely Connected Convolutional Networks.

[3] Convolutional Networks for Biomedical Image Segmentation.

[4] ImageNet Classification with Deep Convolutional Neural Networks.