Discriminatively Learned Hierarchical Rank Pooling Networks

TL;DR

This paper introduces discriminative and hierarchical rank pooling methods for temporal video encoding, enabling end-to-end training of action recognition models that leverage CNN features for improved performance.

Contribution

The paper proposes novel discriminative and hierarchical rank pooling techniques that enhance temporal encoding for action recognition, allowing end-to-end training with CNNs without additional parameters.

Findings

Achieved 76.7% mAP on Hollywood2

Achieved 69.4% on HMDB51

Achieved 93.6% on UCF101

Abstract

In this work, we present novel temporal encoding methods for action and activity classification by extending the unsupervised rank pooling temporal encoding method in two ways. First, we present "discriminative rank pooling" in which the shared weights of our video representation and the parameters of the action classifiers are estimated jointly for a given training dataset of labelled vector sequences using a bilevel optimization formulation of the learning problem. When the frame level features vectors are obtained from a convolutional neural network (CNN), we rank pool the network activations and jointly estimate all parameters of the model, including CNN filters and fully-connected weights, in an end-to-end manner which we coined as "end-to-end trainable rank pooled CNN". Importantly, this model can make use of any existing convolutional neural network architecture (e.g., AlexNet or VGG) without modification or introduction of additional parameters. Then, we extend rank pooling to a high capacity video representation, called "hierarchical rank pooling". Hierarchical rank pooling consists of a network of rank pooling functions, which encode temporal semantics over arbitrary long video clips based on rich frame level features. By stacking non-linear feature functions and temporal sub-sequence encoders one on top of the other, we build a high capacity encoding network of the dynamic behaviour of the video. The resulting video representation is a fixed-length feature vector describing the entire video clip that can be used as input to standard machine learning classifiers. We demonstrate our approach on the task of action and activity recognition. Obtained results are comparable to state-of-the-art methods on three important activity recognition benchmarks with classification performance of 76.7% mAP on Hollywood2, 69.4% on HMDB51, and 93.6% on UCF101.