Improving Weakly-Supervised Object Localization By Micro-Annotation

TL;DR

This paper introduces a novel approach to improve weakly-supervised object localization by incorporating minimal additional annotations through clustering deep network representations, significantly enhancing performance on challenging datasets.

Contribution

The paper proposes a micro-annotation method that leverages clustering of deep features to correct localization errors in weakly-supervised learning.

Findings

Improved bounding box detection on ILSVRC2014 dataset.

Enhanced semantic segmentation on PASCAL VOC2012.

Significant performance gains with minimal additional annotation.

Abstract

Weakly-supervised object localization methods tend to fail for object classes that consistently co-occur with the same background elements, e.g. trains on tracks. We propose a method to overcome these failures by adding a very small amount of model-specific additional annotation. The main idea is to cluster a deep network's mid-level representations and assign object or distractor labels to each cluster. Experiments show substantially improved localization results on the challenging ILSVC2014 dataset for bounding box detection and the PASCAL VOC2012 dataset for semantic segmentation.