{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/spottune-transfer-learning-through-adaptive","title":"SpotTune: Transfer Learning through Adaptive Fine-tuning","arxiv_id":"1811.08737","date":"2018-11-21","proceeding":"CVPR 2019 6","authors":["Yunhui Guo","Honghui Shi","Abhishek Kumar","Kristen Grauman","Tajana Rosing","Rogerio Feris"],"abstract":"Transfer learning, which allows a source task to affect the inductive bias of\nthe target task, is widely used in computer vision. The typical way of\nconducting transfer learning with deep neural networks is to fine-tune a model\npre-trained on the source task using data from the target task. In this paper,\nwe propose an adaptive fine-tuning approach, called SpotTune, which finds the\noptimal fine-tuning strategy per instance for the target data. In SpotTune,\ngiven an image from the target task, a policy network is used to make routing\ndecisions on whether to pass the image through the fine-tuned layers or the\npre-trained layers. We conduct extensive experiments to demonstrate the\neffectiveness of the proposed approach. Our method outperforms the traditional\nfine-tuning approach on 12 out of 14 standard datasets.We also compare SpotTune\nwith other state-of-the-art fine-tuning strategies, showing superior\nperformance. On the Visual Decathlon datasets, our method achieves the highest\nscore across the board without bells and whistles.","url_abs":"http://arxiv.org/abs/1811.08737v1","url_pdf":"http://arxiv.org/pdf/1811.08737v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"spottune-transfer-learning-through-adaptive","repo_url":"https://github.com/YuWang24/MultiTune","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"spottune-transfer-learning-through-adaptive","repo_url":"https://github.com/gyhui14/spottune","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"spottune-transfer-learning-through-adaptive","repo_url":"https://github.com/timmywanttolearn/fintune","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"inductive-bias","task_name":"Inductive Bias"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1811.08737","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}