CLIP Adaptation by Intra-modal Overlap Reduction

TL;DR

This paper analyzes the intra-modal overlap in CLIP's image embeddings and proposes a lightweight adapter to reduce this overlap, leading to improved few-shot classification accuracy, robustness, and feature discriminability.

Contribution

It introduces a novel intra-modal overlap reduction method via a lightweight adapter, enhancing CLIP's few-shot classification performance and robustness.

Findings

Reduced intra-modal overlap improves classification accuracy

Enhanced robustness to distribution shifts

Features become more discriminative for downstream tasks

Abstract

Numerous methods have been proposed to adapt a pre-trained foundational CLIP model for few-shot classification. As CLIP is trained on a large corpus, it generalises well through adaptation to few-shot classification. In this work, we analyse the intra-modal overlap in image space in terms of embedding representation. Our analysis shows that, due to contrastive learning, embeddings from CLIP model exhibit high cosine similarity distribution overlap in the image space between paired and unpaired examples affecting the performance of few-shot training-free classification methods which rely on similarity in the image space for their predictions. To tackle intra-modal overlap we propose to train a lightweight adapter on a generic set of samples from the Google Open Images dataset demonstrating that this improves accuracy for few-shot training-free classification. We validate our contribution through extensive empirical analysis and demonstrate that reducing the intra-modal overlap leads to a) improved performance on a number of standard datasets, b) increased robustness to distribution shift and c) higher feature variance rendering the features more discriminative for downstream tasks.