CountCLIP -- [Re] Teaching CLIP to Count to Ten

TL;DR

This paper reproduces and evaluates CountCLIP, a method that fine-tunes CLIP models to enhance zero-shot counting accuracy without sacrificing classification performance, demonstrating improved quantitative understanding of objects.

Contribution

It provides a reproducibility study of CountCLIP, confirming its effectiveness and offering an accessible implementation for further research.

Findings

Improved zero-shot counting accuracy on a subset of data

Maintained zero-shot classification performance

Reproducibility of CountCLIP's results confirmed

Abstract

Large vision-language models (VLMs) are shown to learn rich joint image-text representations enabling high performances in relevant downstream tasks. However, they fail to showcase their quantitative understanding of objects, and they lack good counting-aware representation. This paper conducts a reproducibility study of 'Teaching CLIP to Count to Ten' (Paiss et al., 2023), which presents a method to finetune a CLIP model (Radford et al., 2021) to improve zero-shot counting accuracy in an image while maintaining the performance for zero-shot classification by introducing a counting-contrastive loss term. We improve the model's performance on a smaller subset of their training data with lower computational resources. We verify these claims by reproducing their study with our own code. The implementation can be found at https://github.com/SforAiDl/CountCLIP.