{"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/training-a-fully-convolutional-neural-network","title":"Training a Fully Convolutional Neural Network to Route Integrated Circuits","arxiv_id":"1706.08948","date":"2017-06-27","proceeding":null,"authors":["Sambhav R. Jain","Kye Okabe"],"abstract":"We present a deep, fully convolutional neural network that learns to route a\ncircuit layout net with appropriate choice of metal tracks and wire class\ncombinations. Inputs to the network are the encoded layouts containing spatial\nlocation of pins to be routed. After 15 fully convolutional stages followed by\na score comparator, the network outputs 8 layout layers (corresponding to 4\nroute layers, 3 via layers and an identity-mapped pin layer) which are then\ndecoded to obtain the routed layouts. We formulate this as a binary\nsegmentation problem on a per-pixel per-layer basis, where the network is\ntrained to correctly classify pixels in each layout layer to be 'on' or 'off'.\nTo demonstrate learnability of layout design rules, we train the network on a\ndataset of 50,000 train and 10,000 validation samples that we generate based on\ncertain pre-defined layout constraints. Precision, recall and $F_1$ score\nmetrics are used to track the training progress. Our network achieves\n$F_1\\approx97\\%$ on the train set and $F_1\\approx92\\%$ on the validation set.\nWe use PyTorch for implementing our model. Code is made publicly available at\nhttps://github.com/sjain-stanford/deep-route .","url_abs":"http://arxiv.org/abs/1706.08948v2","url_pdf":"http://arxiv.org/pdf/1706.08948v2.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":"training-a-fully-convolutional-neural-network","repo_url":"https://github.com/sjain-stanford/deep-route","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"layout-design","task_name":"Layout Design"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}