eprintid: 10199672 rev_number: 7 eprint_status: archive userid: 699 dir: disk0/10/19/96/72 datestamp: 2024-11-07 16:03:41 lastmod: 2024-11-07 16:03:41 status_changed: 2024-11-07 16:03:41 type: proceedings_section metadata_visibility: show sword_depositor: 699 creators_name: Feng, Chen creators_name: Tzimiropoulos, Georgios creators_name: Patras, Ioannis title: CLIPCleaner: Cleaning Noisy Labels with CLIP ispublished: pub divisions: UCL divisions: B04 divisions: F46 keywords: Sample selection, Noisy Labels, CLIP note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions. abstract: Learning with Noisy labels (LNL) poses a significant challenge for the Machine Learning community. Some of the most widely used approaches that select as clean samples for which the model itself (the in-training model) has high confidence, e.g., 'small loss', can suffer from the so called 'self-confirmation' bias. This bias arises because the in-training model, is at least partially trained on the noisy labels. Furthermore, in the classification case, an additional challenge arises because some of the label noise is between classes that are visually very similar (`hard noise'). This paper addresses these challenges by proposing a method (CLIPCleaner) that leverages CLIP, a powerful Vision-Language (VL) model for constructing a zero-shot classifier for efficient, offline, clean sample selection. This has the advantage that the sample selection is decoupled from the in-training model and that the sample selection is aware of the semantic and visual similarities between the classes due to the way that CLIP is trained. We provide theoretical justifications and empirical evidence to demonstrate the advantages of CLIP for LNL compared to conventional pre-trained models. Compared to current methods that combine iterative sample selection with various techniques, CLIPCleaner offers a simple, single-step approach that achieves competitive or superior performance on benchmark datasets. To the best of our knowledge, this is the first time a VL model has been used for sample selection to address the problem of Learning with Noisy Labels (LNL), highlighting their potential in the domain. date: 2024-10-28 date_type: published publisher: ACM official_url: https://doi.org/10.1145/3664647.3680664 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 2332666 doi: 10.1145/3664647.3680664 lyricists_name: Feng, Chen lyricists_id: CFENA90 actors_name: Flynn, Bernadette actors_id: BFFLY94 actors_role: owner full_text_status: public pres_type: paper publication: Proceedings of the 32nd ACM International Conference on Multimedia pagerange: 876-885 event_title: MM '24: The 32nd ACM International Conference on Multimedia book_title: Proceedings of the 32nd ACM International Conference on Multimedia citation: Feng, Chen; Tzimiropoulos, Georgios; Patras, Ioannis; (2024) CLIPCleaner: Cleaning Noisy Labels with CLIP. In: Proceedings of the 32nd ACM International Conference on Multimedia. (pp. pp. 876-885). ACM Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10199672/1/ACMMM2024_CLIPCleaner_camera_ready.pdf