Evaluating Weakly Supervised Object Localization Methods Right

TL;DR

This paper critically evaluates recent weakly-supervised object localization methods, revealing they have not significantly improved over baseline methods when using a new evaluation protocol that limits supervision, and discusses future research directions.

Contribution

It introduces a new evaluation protocol for WSOL that limits full supervision to a small held-out set, revealing current methods' limited progress and suggesting future research directions.

Findings

Recent WSOL methods do not outperform the CAM baseline under the new protocol.

Existing WSOL methods have not surpassed the few-shot learning baseline.

The WSOL task may be ill-posed with only image-level labels.

Abstract

Weakly-supervised object localization (WSOL) has gained popularity over the last years for its promise to train localization models with only image-level labels. Since the seminal WSOL work of class activation mapping (CAM), the field has focused on how to expand the attention regions to cover objects more broadly and localize them better. However, these strategies rely on full localization supervision to validate hyperparameters and for model selection, which is in principle prohibited under the WSOL setup. In this paper, we argue that WSOL task is ill-posed with only image-level labels, and propose a new evaluation protocol where full supervision is limited to only a small held-out set not overlapping with the test set. We observe that, under our protocol, the five most recent WSOL methods have not made a major improvement over the CAM baseline. Moreover, we report that existing WSOL methods have not reached the few-shot learning baseline, where the full-supervision at validation time is used for model training instead. Based on our findings, we discuss some future directions for WSOL.