{"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/usb-universal-scale-object-detection","title":"USB: Universal-Scale Object Detection Benchmark","arxiv_id":"2103.14027","date":"2021-03-25","proceeding":null,"authors":["Yosuke Shinya"],"abstract":"Benchmarks, such as COCO, play a crucial role in object detection. However, existing benchmarks are insufficient in scale variation, and their protocols are inadequate for fair comparison. In this paper, we introduce the Universal-Scale object detection Benchmark (USB). USB has variations in object scales and image domains by incorporating COCO with the recently proposed Waymo Open Dataset and Manga109-s dataset. To enable fair comparison and inclusive research, we propose training and evaluation protocols. They have multiple divisions for training epochs and evaluation image resolutions, like weight classes in sports, and compatibility across training protocols, like the backward compatibility of the Universal Serial Bus. Specifically, we request participants to report results with not only higher protocols (longer training) but also lower protocols (shorter training). Using the proposed benchmark and protocols, we conducted extensive experiments using 15 methods and found weaknesses of existing COCO-biased methods. The code is available at https://github.com/shinya7y/UniverseNet .","url_abs":"https://arxiv.org/abs/2103.14027v3","url_pdf":"https://arxiv.org/pdf/2103.14027v3.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":"usb-universal-scale-object-detection","repo_url":"https://github.com/shinya7y/UniverseNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"real-time-object-detection","task_name":"Real-Time Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"cspdarknet53","method_name":"CSPDarknet53"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"yolox","method_name":"YOLOX"}],"datasets_introduced":[{"slug":"usb","name":"USB","full_name":"Universal-Scale Object Detection Benchmark"}],"methods_introduced":[],"results":[{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"UniverseNet-20.08d (Res2Net-101, DCN, multi-scale)","rank_in_archive_order":60,"of":220,"metrics":{"AP50":"70.8","AP75":"58.9","APL":"68.1","APM":"57.5","APS":"36.9","box AP":"53.5"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"UniverseNet-20.08d (Res2Net-101, DCN, single-scale)","rank_in_archive_order":76,"of":220,"metrics":{"AP50":"69.5","AP75":"55.4","APL":"65.8","APM":"55.5","APS":"33.5","box AP":"50.9"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"UniverseNet-20.08 (Res2Net-50, DCN, single-scale)","rank_in_archive_order":89,"of":220,"metrics":{"AP50":"67.0","AP75":"52.6","APL":"62.7","APM":"52.7","APS":"30.6","box AP":"48.5"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco","task":"Object Detection","dataset":"COCO test-dev","model":"UniverseNet-20.08d (Res2Net-101, DCN, multi-scale)","rank_in_archive_order":57,"of":225,"metrics":{"AP50":"71.6","AP75":"59.9","APL":"67.4","APM":"57.2","APS":"35.8","box mAP":"54.1"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco","task":"Object Detection","dataset":"COCO test-dev","model":"UniverseNet-20.08d (Res2Net-101, DCN, single-scale)","rank_in_archive_order":81,"of":225,"metrics":{"AP50":"70.0","AP75":"55.8","APL":"64.9","APM":"55.3","APS":"31.7","box mAP":"51.3"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco","task":"Object Detection","dataset":"COCO test-dev","model":"UniverseNet-20.08 (Res2Net-50, DCN, single-scale)","rank_in_archive_order":102,"of":225,"metrics":{"AP50":"67.5","AP75":"53.0","APL":"61.1","APM":"52.3","APS":"30.1","box mAP":"48.8"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-o","task":"Object Detection","dataset":"COCO-O","model":"UniverseNet\n(R2-101-DCN)","rank_in_archive_order":28,"of":45,"metrics":{"Average mAP":"24.8"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-o","task":"Object Detection","dataset":"COCO-O","model":"UniverseNet (R2-101-DCN)","rank_in_archive_order":44,"of":45,"metrics":{"Effective Robustness":"1.86"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"YOLOX-L","rank_in_archive_order":1,"of":15,"metrics":{"COCO-style AP":"70.2"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"UniverseNet-20.08","rank_in_archive_order":2,"of":15,"metrics":{"COCO-style AP":"69.9"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"UniverseNet","rank_in_archive_order":3,"of":15,"metrics":{"COCO-style AP":"68.9"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"ATSS+DyHead","rank_in_archive_order":4,"of":15,"metrics":{"COCO-style AP":"67.9"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"Cascade R-CNN","rank_in_archive_order":5,"of":15,"metrics":{"COCO-style AP":"67.6"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"ATSS (ConvNeXt-T)","rank_in_archive_order":6,"of":15,"metrics":{"COCO-style AP":"67.4"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"GFL","rank_in_archive_order":7,"of":15,"metrics":{"COCO-style AP":"67.3"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"ATSS+SEPC","rank_in_archive_order":8,"of":15,"metrics":{"COCO-style AP":"67.1"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"ATSS","rank_in_archive_order":9,"of":15,"metrics":{"COCO-style AP":"66.5"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"ATSS (Swin-T)","rank_in_archive_order":10,"of":15,"metrics":{"COCO-style AP":"66.2"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"Faster R-CNN","rank_in_archive_order":11,"of":15,"metrics":{"COCO-style AP":"65.8"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"RetinaNet","rank_in_archive_order":12,"of":15,"metrics":{"COCO-style AP":"65.3"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"Deformable DETR","rank_in_archive_order":13,"of":15,"metrics":{"COCO-style AP":"64.1"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"Sparse R-CNN","rank_in_archive_order":14,"of":15,"metrics":{"COCO-style AP":"63.1"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-manga109-s-15test","task":"Object Detection","dataset":"Manga109-s 15test","model":"DETR","rank_in_archive_order":15,"of":15,"metrics":{"COCO-style AP":"31.2"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"UniverseNet-20.08","rank_in_archive_order":1,"of":13,"metrics":{"mCAP":"52.1"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"UniverseNet","rank_in_archive_order":2,"of":13,"metrics":{"mCAP":"51.4"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"ATSS (ConvNeXt-T)","rank_in_archive_order":3,"of":13,"metrics":{"mCAP":"50.4"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"YOLOX-L","rank_in_archive_order":4,"of":13,"metrics":{"mCAP":"49.6"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"ATSS+DyHead","rank_in_archive_order":5,"of":13,"metrics":{"mCAP":"49.4"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"ATSS (Swin-T)","rank_in_archive_order":6,"of":13,"metrics":{"mCAP":"49.0"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"Cascade R-CNN","rank_in_archive_order":7,"of":13,"metrics":{"mCAP":"48.1"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"GFL","rank_in_archive_order":8,"of":13,"metrics":{"mCAP":"47.7"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"ATSS","rank_in_archive_order":9,"of":13,"metrics":{"mCAP":"47.1"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"Faster R-CNN","rank_in_archive_order":10,"of":13,"metrics":{"mCAP":"45.9"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"RetinaNet","rank_in_archive_order":11,"of":13,"metrics":{"mCAP":"44.8"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"Sparse R-CNN","rank_in_archive_order":12,"of":13,"metrics":{"mCAP":"44.6"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-usb-standard-usb-1-0","task":"Object Detection","dataset":"USB (Standard USB 1.0 protocol)","model":"DETR","rank_in_archive_order":13,"of":13,"metrics":{"mCAP":"23.7"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"YOLOX-L","rank_in_archive_order":1,"of":15,"metrics":{"COCO-style AP":"41.6"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"UniverseNet-20.08","rank_in_archive_order":2,"of":15,"metrics":{"COCO-style AP":"39.0"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"UniverseNet","rank_in_archive_order":3,"of":15,"metrics":{"COCO-style AP":"38.6"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"ATSS (ConvNeXt-T)","rank_in_archive_order":4,"of":15,"metrics":{"COCO-style AP":"38.3"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"ATSS (Swin-T)","rank_in_archive_order":5,"of":15,"metrics":{"COCO-style AP":"37.2"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"ATSS+DyHead","rank_in_archive_order":6,"of":15,"metrics":{"COCO-style AP":"37.1"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"Cascade R-CNN","rank_in_archive_order":7,"of":15,"metrics":{"COCO-style AP":"36.4"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"GFL","rank_in_archive_order":8,"of":15,"metrics":{"COCO-style AP":"35.7"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"ATSS","rank_in_archive_order":9,"of":15,"metrics":{"COCO-style AP":"35.4"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"ATSS+SEPC","rank_in_archive_order":10,"of":15,"metrics":{"COCO-style AP":"35.0"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"Faster R-CNN","rank_in_archive_order":11,"of":15,"metrics":{"COCO-style AP":"34.5"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"Sparse R-CNN","rank_in_archive_order":12,"of":15,"metrics":{"COCO-style AP":"32.8"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"Deformable DETR","rank_in_archive_order":13,"of":15,"metrics":{"COCO-style AP":"32.7"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"RetinaNet","rank_in_archive_order":14,"of":15,"metrics":{"COCO-style AP":"32.5"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns-1","task":"Object Detection","dataset":"Waymo 2D detection all_ns f0val","model":"DETR","rank_in_archive_order":15,"of":15,"metrics":{"COCO-style AP":"17.8"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-waymo-2d-detection-all-ns","task":"Object Detection","dataset":"Waymo 2D detection all_ns test","model":"UniverseNet","rank_in_archive_order":1,"of":1,"metrics":{"AP/L2":"67.42"},"uses_additional_data":true}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2103.14027","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}