Transductive Zero-Shot and Few-Shot CLIP

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Martin, Ségolène, Huang, Yunshi, Shakeri, Fereshteh, Pesquet, Jean-Christophe, Ayed, Ismail Ben
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913368022974464
author Martin, Ségolène
Huang, Yunshi
Shakeri, Fereshteh
Pesquet, Jean-Christophe
Ayed, Ismail Ben
author_facet Martin, Ségolène
Huang, Yunshi
Shakeri, Fereshteh
Pesquet, Jean-Christophe
Ayed, Ismail Ben
contents Transductive inference has been widely investigated in few-shot image classification, but completely overlooked in the recent, fast growing literature on adapting vision-langage models like CLIP. This paper addresses the transductive zero-shot and few-shot CLIP classification challenge, in which inference is performed jointly across a mini-batch of unlabeled query samples, rather than treating each instance independently. We initially construct informative vision-text probability features, leading to a classification problem on the unit simplex set. Inspired by Expectation-Maximization (EM), our optimization-based classification objective models the data probability distribution for each class using a Dirichlet law. The minimization problem is then tackled with a novel block Majorization-Minimization algorithm, which simultaneously estimates the distribution parameters and class assignments. Extensive numerical experiments on 11 datasets underscore the benefits and efficacy of our batch inference approach.On zero-shot tasks with test batches of 75 samples, our approach yields near 20% improvement in ImageNet accuracy over CLIP's zero-shot performance. Additionally, we outperform state-of-the-art methods in the few-shot setting. The code is available at: https://github.com/SegoleneMartin/transductive-CLIP.
format Preprint
id arxiv_https___arxiv_org_abs_2405_18437
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Transductive Zero-Shot and Few-Shot CLIP
Martin, Ségolène
Huang, Yunshi
Shakeri, Fereshteh
Pesquet, Jean-Christophe
Ayed, Ismail Ben
Computer Vision and Pattern Recognition
Artificial Intelligence
Transductive inference has been widely investigated in few-shot image classification, but completely overlooked in the recent, fast growing literature on adapting vision-langage models like CLIP. This paper addresses the transductive zero-shot and few-shot CLIP classification challenge, in which inference is performed jointly across a mini-batch of unlabeled query samples, rather than treating each instance independently. We initially construct informative vision-text probability features, leading to a classification problem on the unit simplex set. Inspired by Expectation-Maximization (EM), our optimization-based classification objective models the data probability distribution for each class using a Dirichlet law. The minimization problem is then tackled with a novel block Majorization-Minimization algorithm, which simultaneously estimates the distribution parameters and class assignments. Extensive numerical experiments on 11 datasets underscore the benefits and efficacy of our batch inference approach.On zero-shot tasks with test batches of 75 samples, our approach yields near 20% improvement in ImageNet accuracy over CLIP's zero-shot performance. Additionally, we outperform state-of-the-art methods in the few-shot setting. The code is available at: https://github.com/SegoleneMartin/transductive-CLIP.
title Transductive Zero-Shot and Few-Shot CLIP
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2405.18437