{"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/spatial-memory-for-context-reasoning-in","title":"Spatial Memory for Context Reasoning in Object Detection","arxiv_id":"1704.04224","date":"2017-04-13","proceeding":"ICCV 2017 10","authors":["Xinlei Chen","Abhinav Gupta"],"abstract":"Modeling instance-level context and object-object relationships is extremely\nchallenging. It requires reasoning about bounding boxes of different classes,\nlocations \\etc. Above all, instance-level spatial reasoning inherently requires\nmodeling conditional distributions on previous detections. Unfortunately, our\ncurrent object detection systems do not have any {\\bf memory} to remember what\nto condition on! The state-of-the-art object detectors still detect all object\nin parallel followed by non-maximal suppression (NMS). While memory has been\nused for tasks such as captioning, they mostly use image-level memory cells\nwithout capturing the spatial layout. On the other hand, modeling object-object\nrelationships requires {\\bf spatial} reasoning -- not only do we need a memory\nto store the spatial layout, but also a effective reasoning module to extract\nspatial patterns. This paper presents a conceptually simple yet powerful\nsolution -- Spatial Memory Network (SMN), to model the instance-level context\nefficiently and effectively. Our spatial memory essentially assembles object\ninstances back into a pseudo \"image\" representation that is easy to be fed into\nanother ConvNet for object-object context reasoning. This leads to a new\nsequential reasoning architecture where image and memory are processed in\nparallel to obtain detections which update the memory again. We show our SMN\ndirection is promising as it provides 2.2\\% improvement over baseline Faster\nRCNN on the COCO dataset so far.","url_abs":"http://arxiv.org/abs/1704.04224v1","url_pdf":"http://arxiv.org/pdf/1704.04224v1.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":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/endernewton/tf-faster-rcnn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/1512159/tf-faster-rcnn-medico","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/AlphaJia/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/ChestnutLi/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/Kenneth-Wong/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/KyeongmoonKim/sb","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/PengchengAi/tf-faster-rcnn-pcai","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/PythonZhengbinChen/tf-faster-rcnn-cpu","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/achiyaj/vqa-sandbox","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/bareblackfoot/faster-rcnn.selection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/bareblackfoot/lddp-tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/buaaYYC/faster-rcnn-tensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/csmiler/tf-faster-rcnn-cpu","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/daxiapazi/faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/hj963434688/Faster_rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/huan123/py-fatser-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/insigh/Faster-RCNN-Tensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/ithuanhuan/gpu-py3-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/ithuanhuan/py-fatser-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/liuqiang3/faster_rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/mappy/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/mengyingfei/faster-rcnn-tf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/nautilus261/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/neuqgz/modify-faster-rcnn-tf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/piaosonglin1985/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/rickyHong/faster-rcnn-tensorflow-repl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/shikorab/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/tianhai123/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/tigerofmurder/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/vietlinhtspt/NewFasterRCNN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/wangzpeng/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/wennieWN/endernewton_tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/xinpingwang/tf-faster-rcnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/yzgrfsy/tf-fastrcnn-crop","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"spatial-memory-for-context-reasoning-in","repo_url":"https://github.com/zxqcreations/faster-rcnn-tf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"spatial-reasoning","task_name":"Spatial Reasoning"}],"methods":[{"method_slug":"memory-network","method_name":"Memory Network"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1704.04224","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}