LOOC: Localize Overlapping Objects with Count Supervision

TL;DR

LOOC introduces a novel approach to localize overlapping objects in dense scenes using only count supervision, combining semi-supervised pseudo label generation with fully-supervised training to improve localization and counting accuracy.

Contribution

The paper proposes LOOC, a new method that localizes overlapping objects with count supervision by iteratively generating pseudo labels and refining localization.

Findings

LOOC sets a new baseline for localization with count supervision.

LOOC outperforms state-of-the-art counting methods using only count data.

The method effectively handles dense scenes with overlapping objects.

Abstract

Acquiring count annotations generally requires less human effort than point-level and bounding box annotations. Thus, we propose the novel problem setup of localizing objects in dense scenes under this weaker supervision. We propose LOOC, a method to Localize Overlapping Objects with Count supervision. We train LOOC by alternating between two stages. In the first stage, LOOC learns to generate pseudo point-level annotations in a semi-supervised manner. In the second stage, LOOC uses a fully-supervised localization method that trains on these pseudo labels. The localization method is used to progressively improve the quality of the pseudo labels. We conducted experiments on popular counting datasets. For localization, LOOC achieves a strong new baseline in the novel problem setup where only count supervision is available. For counting, LOOC outperforms current state-of-the-art methods that only use count as their supervision. Code is available at: https://github.com/ElementAI/looc.