{"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/towards-interpretable-r-cnn-by-unfolding","title":"Towards Interpretable R-CNN by Unfolding Latent Structures","arxiv_id":"1711.05226","date":"2017-11-14","proceeding":null,"authors":["Tianfu Wu","Wei Sun","Xilai Li","Xi Song","Bo Li"],"abstract":"This paper first proposes a method of formulating model interpretability in\nvisual understanding tasks based on the idea of unfolding latent structures. It\nthen presents a case study in object detection using popular two-stage\nregion-based convolutional network (i.e., R-CNN) detection systems. We focus on\nweakly-supervised extractive rationale generation, that is learning to unfold\nlatent discriminative part configurations of object instances automatically and\nsimultaneously in detection without using any supervision for part\nconfigurations. We utilize a top-down hierarchical and compositional grammar\nmodel embedded in a directed acyclic AND-OR Graph (AOG) to explore and unfold\nthe space of latent part configurations of regions of interest (RoIs). We\npropose an AOGParsing operator to substitute the RoIPooling operator widely\nused in R-CNN. In detection, a bounding box is interpreted by the best parse\ntree derived from the AOG on-the-fly, which is treated as the qualitatively\nextractive rationale generated for interpreting detection. We propose a\nfolding-unfolding method to train the AOG and convolutional networks\nend-to-end. In experiments, we build on R-FCN and test our method on the PASCAL\nVOC 2007 and 2012 datasets. We show that the method can unfold promising latent\nstructures without hurting the performance.","url_abs":"http://arxiv.org/abs/1711.05226v2","url_pdf":"http://arxiv.org/pdf/1711.05226v2.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":"towards-interpretable-r-cnn-by-unfolding","repo_url":"https://github.com/msracver/Deformable-ConvNets","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mxnet","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"interpretability","method_name":"Interpretability"},{"method_slug":"position-sensitive-roi-pooling","method_name":"Position-Sensitive RoI Pooling"},{"method_slug":"r-fcn","method_name":"R-FCN"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}