VIDEO CLASSIFICATION WITH CONVOLUTIONAL NEURAL NETWORKS

Abstract

Convolutional Neural Networks (CNNs) have been established as a powerful class of models for image recognition problems. 
Encouraged by these results, we pro-vide an extensive empirical evaluation of CNNs on large- scale video classification using 
a new dataset of 1 million YouTube videos belonging to 487 classes. We study multiple approaches for extending the 
connectivity of a CNN in time domain to take advantage of local spatial-temporal information and suggest a multi resolution, 
for related architecture as a promising way of speeding up the training. Our best spatio temporal networks display significant 
performance improvements compared to strong feature-based baselines (55.3% to 63.9%), but only a surprisingly mod- est 
improvement compared to single-frame models (59.3% to60.9%).We further study the generalization performance of our best 
model by retraining the top layers on the UCF- 101 Action Recognition dataset and observe significant performance 
improvements compared to theUCF-101 baseline model (63.3% up from 43.9%).