C2D(CVPR-2014)
paper : Large-scale Video Classification with Convolutional Neural Networks
idea :
除了Slow Fusion之外,其他fusion的方式,本質上都屬於C2D。
Architecture :
Note:
a context stream that models low-resolution image
a fovea stream that processes high-resolution center crop.
採用Multiresolution CNNs的motivation:加快訓練速度(通過減小輸入frame的size)