R-C3D: Region Convolutional 3D Network for Temporal Activity Detection

TL;DR

R-C3D is an end-to-end 3D convolutional network that efficiently detects and classifies activities in untrimmed videos, achieving state-of-the-art results and high speed across multiple datasets.

Contribution

The paper introduces R-C3D, a novel fully convolutional 3D network for activity detection that shares features for proposal and classification, enabling fast and accurate detection.

Findings

Achieves real-time detection at 569 fps.

Outperforms existing methods on THUMOS'14.

Effective across multiple datasets like ActivityNet and Charades.

Abstract

We address the problem of activity detection in continuous, untrimmed video streams. This is a difficult task that requires extracting meaningful spatio-temporal features to capture activities, accurately localizing the start and end times of each activity. We introduce a new model, Region Convolutional 3D Network (R-C3D), which encodes the video streams using a three-dimensional fully convolutional network, then generates candidate temporal regions containing activities, and finally classifies selected regions into specific activities. Computation is saved due to the sharing of convolutional features between the proposal and the classification pipelines. The entire model is trained end-to-end with jointly optimized localization and classification losses. R-C3D is faster than existing methods (569 frames per second on a single Titan X Maxwell GPU) and achieves state-of-the-art results on THUMOS'14. We further demonstrate that our model is a general activity detection framework that does not rely on assumptions about particular dataset properties by evaluating our approach on ActivityNet and Charades. Our code is available at http://ai.bu.edu/r-c3d/.