CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos | Read Paper on Bytez