{"type":"ScholarlyArticle","title":"Brain-Inspired Stepwise Patch Merging for Vision Transformers","authors":[{"name":"Yonghao Yu","slug":"yonghao-yu","given":"Yonghao","family":"Yu","affiliation":""},{"name":"Dongcheng Zhao","slug":"dongcheng-zhao","given":"Dongcheng","family":"Zhao","affiliation":""},{"name":"Guobin Shen","slug":"guobin-shen","given":"Guobin","family":"Shen","affiliation":""},{"name":"Yiting Dong","slug":"yiting-dong","given":"Yiting","family":"Dong","affiliation":""},{"name":"Yi Zeng","slug":"yi-zeng","given":"Yi","family":"Zeng","affiliation":""}],"abstract":"The hierarchical architecture has become a mainstream design paradigm for Vision Transformers (ViTs), with Patch Merging serving as the pivotal component that transforms a columnar architecture into a hierarchical one. Drawing inspiration from the brain's ability to integrate global and local information for comprehensive visual understanding, we propose Stepwise Patch Merging (SPM), which enhances the subsequent attention mechanism's ability to 'see' better. SPM consists of Multi-Scale Aggregation (MSA) and Guided Local Enhancement (GLE) striking a proper balance between long-range dependency modeling and local feature enhancement. Extensive experiments conducted on benchmark datasets, including ImageNet-1K, COCO, and ADE20K, demonstrate that SPM significantly improves the performance of various models, particularly in dense prediction tasks such as object detection and semantic segmentation. Meanwhile, experiments show that combining SPM with different backbones can further improve performance. The code has been released at https://github.com/Yonghao-Yu/StepwisePatchMerging.","doi":"10.24963/ijcai.2025/257","url":"https://ijcai.org/proceedings/2025/257","pdf":"https://ijcai.org/proceedings/2025/0257.pdf","markdown":"https://ijcai.org/proceedings/2025/257.md","bibtex":"https://ijcai.org/proceedings/2025/bibtex/257","volume":{"slug":"2025","title":"Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2025","url":"https://ijcai.org/proceedings/2025/","year":2025,"location":"Montreal, Canada","dates":"16-22 August 2025","editors":"James Kwok","isbn":"978-1-956792-06-5","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2025"},"section":"Computer Vision","track":"Main Track","number":257,"pages":{"first":2305,"last":2313},"keywords":["Humans and AI: HAI: Cognitive modeling","Humans and AI: HAI: Cognitive systems","Computer Vision: CV: Recognition (object detection, categorization)","Computer Vision: CV: Representation learning"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}