{"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/weakly-supervised-dcnn-for-rgb-d-object","title":"Weakly-supervised DCNN for RGB-D Object Recognition in Real-World Applications Which Lack Large-scale Annotated Training Data","arxiv_id":"1703.06370","date":"2017-03-19","proceeding":null,"authors":["Li Sun","Cheng Zhao","Rustam Stolkin"],"abstract":"This paper addresses the problem of RGBD object recognition in real-world\napplications, where large amounts of annotated training data are typically\nunavailable. To overcome this problem, we propose a novel, weakly-supervised\nlearning architecture (DCNN-GPC) which combines parametric models (a pair of\nDeep Convolutional Neural Networks (DCNN) for RGB and D modalities) with\nnon-parametric models (Gaussian Process Classification). Our system is\ninitially trained using a small amount of labeled data, and then automatically\nprop- agates labels to large-scale unlabeled data. We first run 3D- based\nobjectness detection on RGBD videos to acquire many unlabeled object proposals,\nand then employ DCNN-GPC to label them. As a result, our multi-modal DCNN can\nbe trained end-to-end using only a small amount of human annotation. Finally,\nour 3D-based objectness detection and multi-modal DCNN are integrated into a\nreal-time detection and recognition pipeline. In our approach, bounding-box\nannotations are not required and boundary-aware detection is achieved. We also\npropose a novel way to pretrain a DCNN for the depth modality, by training on\nvirtual depth images projected from CAD models. We pretrain our multi-modal\nDCNN on public 3D datasets, achieving performance comparable to\nstate-of-the-art methods on Washington RGBS Dataset. We then finetune the\nnetwork by further training on a small amount of annotated data from our novel\ndataset of industrial objects (nuclear waste simulants). Our weakly supervised\napproach has demonstrated to be highly effective in solving a novel RGBD object\nrecognition application which lacks of human annotations.","url_abs":"http://arxiv.org/abs/1703.06370v1","url_pdf":"http://arxiv.org/pdf/1703.06370v1.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":"weakly-supervised-dcnn-for-rgb-d-object","repo_url":"https://github.com/kevinlisun/romans_stack","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"object-recognition","task_name":"Object Recognition"},{"task_slug":"weakly-supervised-learning","task_name":"Weakly-supervised Learning"}],"methods":[{"method_slug":"dcnn","method_name":"DCNN"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}