Training CLIP models on Data from Scientific Papers
This work addresses the need for better data quality in training CLIP models for researchers and practitioners, but it is incremental as it builds on existing methods with new data.
The paper tackled the problem of improving CLIP model performance by training on higher-quality domain-specific data from scientific papers, finding that small-scale models showed moderate average performance increases.
Contrastive Language-Image Pretraining (CLIP) models are able to capture the semantic relationship of images and texts and have enabled a wide range of applications, from image retrieval to classification. These models are trained with datasets extracted from web crawls, which are of large quantity but limited quality. This paper explores whether limited amounts higher quality data in a specific domain improve the general performance of CLIP models. To this purpose, we extract text-image data from scientific papers hosted in the arXiv and PubMed Central repositories. Experiments on small-scale CLIP models (ViT B/32) show that model performance increases on average, but only moderately. This result indicates that using the data sources considered in the paper to train large-scale CLIP models is a worthwile research direction.