Markerless hand pose estimation via CNN on Keras

This project aims at developing a deep learning model for markerless hand pose estimation based on a variation of ResNet with readout layers that predict the location of a body part. It is implemented on Keras using Tensorflow as backend.

This project is developed by taking as a reference the works done in DeepLabCut, in which transfer learning has been applied in various pose estimation tasks.

Labeling example

Installation and running instructions will follow ASAP along with a detailed report.

Authors

• Nicola De Carli: [email protected]
• Angelica Ginnante: [email protected]
• Enrico Casagrande: [email protected]
• Alberto Ghiotto: [email protected]
• Claudio Curti: [email protected]

A special thanks goes to Dr. Matias Valdenegro-Toro for helping out with the fundamental implementative details in Keras.