The ImageNet Shuffle: Reorganized Pre-training for Video Event Detection

TL;DR

This paper introduces a novel method for reorganizing ImageNet hierarchy to improve deep neural network pre-training for video event detection, leading to state-of-the-art results on TRECVID datasets.

Contribution

It proposes a bottom-up and top-down hierarchy reorganization of ImageNet for better pre-training, enhancing video event detection performance.

Findings

Improved performance over standard pre-training.

Complementary benefits among different reorganizations.

Achieved state-of-the-art results on TRECVID datasets.

Abstract

This paper strives for video event detection using a representation learned from deep convolutional neural networks. Different from the leading approaches, who all learn from the 1,000 classes defined in the ImageNet Large Scale Visual Recognition Challenge, we investigate how to leverage the complete ImageNet hierarchy for pre-training deep networks. To deal with the problems of over-specific classes and classes with few images, we introduce a bottom-up and top-down approach for reorganization of the ImageNet hierarchy based on all its 21,814 classes and more than 14 million images. Experiments on the TRECVID Multimedia Event Detection 2013 and 2015 datasets show that video representations derived from the layers of a deep neural network pre-trained with our reorganized hierarchy i) improves over standard pre-training, ii) is complementary among different reorganizations, iii) maintains the benefits of fusion with other modalities, and iv) leads to state-of-the-art event detection results. The reorganized hierarchies and their derived Caffe models are publicly available at http://tinyurl.com/imagenetshuffle.