Car Object Counting and Position Estimation via Extension of the CLIP-EBC Framework

TL;DR

This paper extends the CLIP-EBC framework from crowd counting to car object counting and position estimation, demonstrating competitive performance and proposing a clustering method for localization.

Contribution

It introduces a novel application of CLIP-EBC to vehicle counting and proposes a clustering approach for position estimation.

Findings

Achieved second-best performance on CARPK dataset

Proposed a K-means weighted clustering method for localization

Demonstrated potential for extending the framework to object localization

Abstract

In this paper, we investigate the applicability of the CLIP-EBC framework, originally designed for crowd counting, to car object counting using the CARPK dataset. Experimental results show that our model achieves second-best performance compared to existing methods. In addition, we propose a K-means weighted clustering method to estimate object positions based on predicted density maps, indicating the framework's potential extension to localization tasks.