Video Classification with Channel-Separated Convolutional Networks

ICCV 2019 Du TranHeng WangLorenzo TorresaniMatt Feiszli

Group convolution has been shown to offer great computational savings in various 2D convolutional architectures for image classification. It is natural to ask: 1) if group convolution can help to alleviate the high computational cost of video classification networks; 2) what factors matter the most in 3D group convolutional networks; and 3) what are good computation/accuracy trade-offs with 3D group convolutional networks... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
COMPARE
Action Classification Kinetics-400 ip-CSN-152 Accuracy 77.0 # 8
Action Classification Kinetics-400 ip-CSN-152 (Sports-1M pretraining) Accuracy 79.2 # 1
Action Recognition In Videos Something-Something V1 ir-CSN-101 Top 1 Accuracy 48.4 # 12
Action Recognition In Videos Something-Something V1 ir-CSN-152 Top 1 Accuracy 49.3 # 9
Action Recognition In Videos Sports-1M ip-CSN-152 Video [email protected] 75.5 # 1
Action Recognition In Videos Sports-1M ip-CSN-152 Video [email protected] 92.8 # 1