Object-Extent Pooling for Weakly Supervised Single-Shot Localization

TL;DR

This paper introduces a real-time, weakly-supervised single-shot object localization method that eliminates region proposals using a novel pooling technique, enabling fast and efficient localization with only image-level labels.

Contribution

It proposes the first weakly-supervised single-shot detector using CAMs and a new SPAM pooling method, achieving real-time performance without region proposals.

Findings

Achieves 35fps inference speed, significantly faster than previous methods.

Uses a novel SPAM pooling layer that balances max and average pooling.

First to eliminate region proposals in weakly-supervised localization.

Abstract

In the face of scarcity in detailed training annotations, the ability to perform object localization tasks in real-time with weak-supervision is very valuable. However, the computational cost of generating and evaluating region proposals is heavy. We adapt the concept of Class Activation Maps (CAM) into the very first weakly-supervised 'single-shot' detector that does not require the use of region proposals. To facilitate this, we propose a novel global pooling technique called Spatial Pyramid Averaged Max (SPAM) pooling for training this CAM-based network for object extent localisation with only weak image-level supervision. We show this global pooling layer possesses a near ideal flow of gradients for extent localization, that offers a good trade-off between the extremes of max and average pooling. Our approach only requires a single network pass and uses a fast-backprojection technique, completely omitting any region proposal steps. To the best of our knowledge, this is the first approach to do so. Due to this, we are able to perform inference in real-time at 35fps, which is an order of magnitude faster than all previous weakly supervised object localization frameworks.