Seminar 2022-12-21

[ArieRS]

Seminar 2022-12-21

Progress

• Reservoir Computing Progress
  • Continuing experiment on WLASL dataset for 100 dataset
    • Class: 100
    • Total Training Video: 1780 videos
    • Total Testing Video: 258 videos
    • FPS: 25
  • Problem: some frames are not extracted

Method	Architecture	Accuracy	Testing Time (Sec)
RC with Multivariate Dataset	450 Reservoir	14.7%	0.032
RC with Multivariate Dataset	800 Reservoir	19%	0.0633
I3D	Inflated 3D ConvNet	65.89%	no information
Pose-TGCN	pose-based temporal graph convolution networks	55.43%	no information

Next Plan

• Creating research posters for symposiums - Deadline 8th December 2022
• Continuing to explore reservoir computing on multivariate dataset
• Implementing VAE or another dimensional reduction method
• Drafting journal or conference paper for reservoir computing
• Trying out existing implementations to gain insight into gesture recognition
• Try to implement Reservoir Computing
• Drafting symposium paper
• Adding literature study