ReCLIP: Refine Contrastive Language Image Pre-Training with Source Free Domain Adaptation
This addresses the challenge of adapting vision-language models to new domains without labeled data, which is incremental as it builds on CLIP.
The paper tackles the problem of domain gaps and cross-modality misalignment when applying CLIP to downstream target domains, proposing ReCLIP, a source-free domain adaptation method that reduces the average error rate from 30.17% to 25.06% on 22 image classification benchmarks.
Large-scale Pre-Training Vision-Language Model such as CLIP has demonstrated outstanding performance in zero-shot classification, e.g. achieving 76.3% top-1 accuracy on ImageNet without seeing any example, which leads to potential benefits to many tasks that have no labeled data. However, while applying CLIP to a downstream target domain, the presence of visual and text domain gaps and cross-modality misalignment can greatly impact the model performance. To address such challenges, we propose ReCLIP, the first source-free domain adaptation method for vision-language models, which does not require any source data or target labeled data. ReCLIP first learns a projection space to mitigate the misaligned visual-text embeddings and learns pseudo labels, and then deploys cross-modality self-training with the pseudo labels, to update visual and text encoders, refine labels and reduce domain gaps and misalignments iteratively. With extensive experiments, we demonstrate ReCLIP reduces the average error rate of CLIP from 30.17% to 25.06% on 22 image classification benchmarks. Code available at https://github.com/michiganleon/ReCLIP_WACV.