{"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/superhuman-accuracy-on-the-snemi3d","title":"Superhuman Accuracy on the SNEMI3D Connectomics Challenge","arxiv_id":"1706.00120","date":"2017-05-31","proceeding":null,"authors":["Kisuk Lee","Jonathan Zung","Peter Li","Viren Jain","H. Sebastian Seung"],"abstract":"For the past decade, convolutional networks have been used for 3D\nreconstruction of neurons from electron microscopic (EM) brain images. Recent\nyears have seen great improvements in accuracy, as evidenced by submissions to\nthe SNEMI3D benchmark challenge. Here we report the first submission to surpass\nthe estimate of human accuracy provided by the SNEMI3D leaderboard. A variant\nof 3D U-Net is trained on a primary task of predicting affinities between\nnearest neighbor voxels, and an auxiliary task of predicting long-range\naffinities. The training data is augmented by simulated image defects. The\nnearest neighbor affinities are used to create an oversegmentation, and then\nsupervoxels are greedily agglomerated based on mean affinity. The resulting\nSNEMI3D score exceeds the estimate of human accuracy by a large margin. While\none should be cautious about extrapolating from the SNEMI3D benchmark to\nreal-world accuracy of large-scale neural circuit reconstruction, our result\ninspires optimism that the goal of full automation may be realizable in the\nfuture.","url_abs":"http://arxiv.org/abs/1706.00120v1","url_pdf":"http://arxiv.org/pdf/1706.00120v1.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":"superhuman-accuracy-on-the-snemi3d","repo_url":"https://github.com/seung-lab/Augmentor","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"superhuman-accuracy-on-the-snemi3d","repo_url":"https://github.com/seung-lab/DataProvider3","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}},{"paper_slug":"superhuman-accuracy-on-the-snemi3d","repo_url":"https://github.com/seung-lab/DeepEM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"superhuman-accuracy-on-the-snemi3d","repo_url":"https://github.com/wolny/pytorch-3dunet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"3d-reconstruction","task_name":"3D Reconstruction"},{"task_slug":"electron-microscopy-image-segmentation","task_name":"Electron Microscopy Image Segmentation"}],"methods":[{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"u-net","method_name":"U-Net"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1706.00120","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}