Don't Stop Learning: Towards Continual Learning for the CLIP Model

Yuxuan Ding,Lingqiao Liu,Chunna Tian,Jingyuan Yang,Haoxuan Ding
DOI: https://doi.org/10.48550/arXiv.2207.09248
2022-07-20
Abstract:The Contrastive Language-Image Pre-training (CLIP) Model is a recently proposed large-scale pre-train model which attracts increasing attention in the computer vision community. Benefiting from its gigantic image-text training set, the CLIP model has learned outstanding capabilities in zero-shot learning and image-text matching. To boost the recognition performance of CLIP on some target visual concepts, it is often desirable to further update the CLIP model by fine-tuning some classes-of-interest on extra training data. This operation, however, raises an important concern: will the update hurt the zero-shot learning or image-text matching capability of the CLIP, i.e., the catastrophic forgetting issue? If yes, could existing continual learning algorithms be adapted to alleviate the risk of catastrophic forgetting? To answer these questions, this work conducts a systemic study on the continual learning issue of the CLIP model. We construct evaluation protocols to measure the impact of fine-tuning updates and explore different ways to upgrade existing continual learning methods to mitigate the forgetting issue of the CLIP model. Our study reveals the particular challenges of CLIP continual learning problem and lays a foundation for further researches. Moreover, we propose a new algorithm, dubbed Learning without Forgetting via Replayed Vocabulary (VR-LwF), which shows exact effectiveness for alleviating the forgetting issue of the CLIP model.
Computer Vision and Pattern Recognition
What problem does this paper attempt to address?
The problem that this paper attempts to solve is that when the CLIP model is fine - tuned for specific categories, it may lead to the decline of its zero - shot learning ability and image - text matching ability, that is, the catastrophic forgetting problem. Specifically, although the CLIP model has strong zero - shot classification and image - text matching abilities through large - scale image - text pair pre - training, when fine - tuning for certain visual concepts to improve recognition performance, this fine - tuning operation may affect the original zero - shot learning and image - text matching abilities of CLIP. Therefore, the paper aims to study how to alleviate this problem through continuous learning algorithms to ensure that CLIP can still maintain its original excellent performance after fine - tuning for specific categories. To address this problem, the authors conducted a systematic study, constructed an evaluation protocol to measure the impact of fine - tuning updates, and explored different methods to upgrade existing continuous learning techniques to mitigate the forgetting problem in the CLIP model. In addition, they also proposed a new algorithm - Vocabulary Replay - based Learning without Forgetting (VR - LwF), which effectively alleviates the forgetting problem of the CLIP model by enforcing the prediction logits of a set of randomly synthesized categories before and after model updates.