Action Classification on Breakfast Dataset using Temporal Convolutional Networks (TCN)
The goal of this project was to classify each segment from the Breakfast dataset to any of the 47 sub-actions (excluding SIL).
I have used two Action Classification models -
- Encoder- Decoder Temporal Convolutional Network (ED-TCN)
- Multi - Stage Temporal Convolutional Network (MS-TCN)
The sole purpose of this project was to explore various TCN models for Action Classification task.
Carreira J, Zisserman A. Quo vadis, action recognition - a new model and the kinetics dataset[C]//proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 6299-6308.