{"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/do-deep-neural-networks-suffer-from-crowding","title":"Do Deep Neural Networks Suffer from Crowding?","arxiv_id":"1706.08616","date":"2017-06-26","proceeding":"NeurIPS 2017 12","authors":["Anna Volokitin","Gemma Roig","Tomaso Poggio"],"abstract":"Crowding is a visual effect suffered by humans, in which an object that can\nbe recognized in isolation can no longer be recognized when other objects,\ncalled flankers, are placed close to it. In this work, we study the effect of\ncrowding in artificial Deep Neural Networks for object recognition. We analyze\nboth standard deep convolutional neural networks (DCNNs) as well as a new\nversion of DCNNs which is 1) multi-scale and 2) with size of the convolution\nfilters change depending on the eccentricity wrt to the center of fixation.\nSuch networks, that we call eccentricity-dependent, are a computational model\nof the feedforward path of the primate visual cortex. Our results reveal that\nthe eccentricity-dependent model, trained on target objects in isolation, can\nrecognize such targets in the presence of flankers, if the targets are near the\ncenter of the image, whereas DCNNs cannot. Also, for all tested networks, when\ntrained on targets in isolation, we find that recognition accuracy of the\nnetworks decreases the closer the flankers are to the target and the more\nflankers there are. We find that visual similarity between the target and\nflankers also plays a role and that pooling in early layers of the network\nleads to more crowding. Additionally, we show that incorporating the flankers\ninto the images of the training set does not improve performance with crowding.","url_abs":"http://arxiv.org/abs/1706.08616v1","url_pdf":"http://arxiv.org/pdf/1706.08616v1.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":"do-deep-neural-networks-suffer-from-crowding","repo_url":"https://github.com/CBMM/eccentricity","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"do-deep-neural-networks-suffer-from-crowding","repo_url":"https://github.com/voanna/eccentricity","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"object-recognition","task_name":"Object Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1706.08616","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}