A Comparative Study on Effects of Original and Pseudo Labels for Weakly Supervised Learning for Car Localization Problem

TL;DR

This paper compares the effects of original and pseudo labels in weakly supervised car localization, demonstrating that an unsupervised approach can outperform weakly supervised methods on the Compcars dataset.

Contribution

It introduces a novel comparison between original and pseudo labels for car localization and proposes an unsupervised learning approach that surpasses weakly supervised methods.

Findings

Unsupervised learning outperforms weakly supervised localization by approximately 6% on the Compcars dataset.

Generated pseudo labels can effectively replace original labels in localization tasks.

Class Activation Mapping combined with morphological edge detection enables effective localization without full supervision.

Abstract

In this study, the effects of different class labels created as a result of multiple conceptual meanings on localization using Weakly Supervised Learning presented on Car Dataset. In addition, the generated labels are included in the comparison, and the solution turned into Unsupervised Learning. This paper investigates multiple setups for car localization in the images with other approaches rather than Supervised Learning. To predict localization labels, Class Activation Mapping (CAM) is implemented and from the results, the bounding boxes are extracted by using morphological edge detection. Besides the original class labels, generated class labels also employed to train CAM on which turn to a solution to Unsupervised Learning example. In the experiments, we first analyze the effects of class labels in Weakly Supervised localization on the Compcars dataset. We then show that the proposed Unsupervised approach outperforms the Weakly Supervised method in this particular dataset by approximately %6.