video-activity-recognition

This is an ongoing project for my undergraduate final year project.

Goals:

• Use trained 3D convolution model to extract video features
• Classify videos using learned parameters for convolving filters and training Dense layer

Final system is able to classify an incoming live stream in real-time to recognize human actions in a shopping mall. Stream only contains a single human at a time and robust to possible variations in the scenario. System is based on 3D convolution layers and reference link is included below.

Many thanks goes to https:/axon-research/c3d-keras for everything...