A new model for video captioning is developed, using a deep three-dimensional Convolutional Neural Network (C3D) as an encoder for videos and a Recurrent Neural Network (RNN) as a decoder for captions. A novel attention mechanism with spatiotemporal alignment is employed to adaptively and sequentially focus on different layers of CNN features (levels of feature “abstraction”), as well as local spatiotemporal regions of the feature maps at each layer. The proposed approach is evaluated on the YouTube2Text benchmark. Experimental results demonstrate quantitatively the effectiveness of our proposed adaptive spatiotemporal feature abstraction for translating videos to sentences with rich semantic structures.
|Original language||English (US)|
|Title of host publication||5th International Conference on Learning Representations, ICLR 2017 - Workshop Track Proceedings|
|Publisher||International Conference on Learning Representations, ICLR|
|State||Published - Jan 1 2019|