{"id":19204302,"url":"https://github.com/wangzheallen/vsad","last_synced_at":"2026-06-24T10:30:20.646Z","repository":{"id":80125559,"uuid":"66529849","full_name":"wangzheallen/vsad","owner":"wangzheallen","description":"this is the code release for ''Weakly Supervised PatchNets: Describing and Aggregating Local Patches for Scene Recognition''","archived":false,"fork":false,"pushed_at":"2018-02-27T22:35:33.000Z","size":99138,"stargazers_count":43,"open_issues_count":3,"forks_count":17,"subscribers_count":7,"default_branch":"master","last_synced_at":"2025-01-04T14:42:03.662Z","etag":null,"topics":["encoding","feature","scene-recognition"],"latest_commit_sha":null,"homepage":null,"language":"Matlab","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/wangzheallen.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2016-08-25T06:07:17.000Z","updated_at":"2024-02-02T04:43:14.000Z","dependencies_parsed_at":"2023-09-21T03:16:35.067Z","dependency_job_id":null,"html_url":"https://github.com/wangzheallen/vsad","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/wangzheallen%2Fvsad","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/wangzheallen%2Fvsad/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/wangzheallen%2Fvsad/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/wangzheallen%2Fvsad/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/wangzheallen","download_url":"https://codeload.github.com/wangzheallen/vsad/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":240275908,"owners_count":19775615,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["encoding","feature","scene-recognition"],"created_at":"2024-11-09T13:07:08.545Z","updated_at":"2026-06-24T10:30:20.588Z","avatar_url":"https://github.com/wangzheallen.png","language":"Matlab","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cp\u003ethis is the release code for \u003cstrong\u003eWeakly Supervised PatchNets: Describing and Aggregating Local Patches for Scene Recognition\u003c/strong\u003e:\u003c/p\u003e\n\n\u003cpre\u003e\u003ccode\u003eWeakly Supervised PatchNets: Describing and Aggregating Local Patches for Scene Recognition\nZhe Wang, Limin Wang, Yali Wang, Bowen Zhang, and Yu Qiao\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eThe performance is as below: \u003c/p\u003e \n\u003ctable\u003e\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"center\"\u003eacc\u003c/th\u003e\n\u003cth align=\"center\"\u003eMIT_indoor\u003c/th\u003e\n\u003cth align=\"center\"\u003eSUN397\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003emean:\u003c/td\u003e\n\u003ctd align=\"center\"\u003e78.5\u003c/td\u003e\n\u003ctd align=\"center\"\u003e63.5\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003eVLAD:\u003c/td\u003e\n\u003ctd align=\"center\"\u003e83.9\u003c/td\u003e\n\u003ctd align=\"center\"\u003e70.1\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003eFV\u003c/td\u003e\n\u003ctd align=\"center\"\u003e83.6\u003c/td\u003e\n\u003ctd align=\"center\"\u003e69.0\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003eVSAD\u003c/td\u003e\n\u003ctd align=\"center\"\u003e84.9\u003c/td\u003e\n\u003ctd align=\"center\"\u003e71.7\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\u003c/table\u003e\n\n\u003cp\u003eNote: The encoding method based on our \u003cstrong\u003escene_patchenet\u003c/strong\u003e feature surpass human performance on sun397(68.5%).\u003c/p\u003e\n\u003ch4\u003eFeature\u003c/h4\u003e\nwe released the concise and effective feature for MIT indoor feature, it is notated as \u003cstrong\u003ehybrid_PatchNet+VSAD\u003c/strong\u003e in the paper which obtains \u003cstrong\u003e86.1\u003c/strong\u003e accuracy. You can use it as baseline or as complementary feature for further study.\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"http://mmlab.siat.ac.cn/mit_hybrid_vsad.mat\"\u003emit_hybrid_vsad.mat\u003c/a\u003e\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"https://drive.google.com/open?id=1i3uwjWBeO2ke9Ikai_XpQPHjpXEnuH9h\"\u003emit_hybrid_vsad.mat, Google_drive\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003ctable\u003e\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"center\"\u003eacc on MIT\u003c/th\u003e\n\u003cth align=\"center\"\u003edimension\u003c/th\u003e\n\u003cth align=\"center\"\u003estorage\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003e86.1\u003c/td\u003e\n\u003ctd align=\"center\"\u003e100*256*2*2\u003c/td\u003e\n\u003ctd align=\"center\"\u003e1.9G\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\u003c/table\u003e\n\n\n\u003ch4\u003eModel\u003c/h4\u003e\n\u003cp\u003eOur trained \u003cstrong\u003escene_patchnet\u003c/strong\u003e and \u003cstrong\u003eobject_patchenet\u003c/strong\u003e, the model is based on cudnn_v4, if your system is based on cudnn_v5, you can use the code below cudnn_v4 to cudnn_v5: https://github.com/yjxiong/caffe/blob/action_recog/python/bn_convert_style.py \u003c/p\u003e \n\n\u003ctable\u003e\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"center\"\u003eacc\u003c/th\u003e\n\u003cth align=\"center\"\u003eTop5\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003eObject_patchnet_on_ImageNet:\u003c/td\u003e\n\u003ctd align=\"center\"\u003e85.3\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"center\"\u003eScene_patchnet_on_Places205:\u003c/td\u003e\n\u003ctd align=\"center\"\u003e82.7\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\u003c/table\u003e\n\n\u003cp\u003eThey both take 128 * 128 patches as input.\u003c/p\u003e\n\n\u003ch4\u003eCode\u003c/h4\u003e \n\u003cul\u003e\n\u003cli\u003emit_hybrid_vsad.mat     -- you can use this feature as your baseline or to concatenate for further study, it is only 100*256*2*2 dimensions while performs \u003cstrong\u003e86.1\u003c/strong\u003e acc on MIT indoor, you can download from \u003ca href=\"http://mmlab.siat.ac.cn/mit_hybrid_vsad.mat\"\u003emit_hybrid_vsad.mat\u003c/a\u003e\u003c/li\u003e\n\u003cli\u003eextracting_feature_exmaple.m     -- you can use this code as template to extract scene_patchnet_feature or object_patchnet_probability, for scene_patchnet_feature it is global average pool feature and for for object_patchnet_probability it is fully connnect feature with softmax function\u003c/li\u003e\n\u003cli\u003efor_encoder_scene67.mat          -- serve as assist to your handle on MIT_indoor dataset, from vl_feat\u003c/li\u003e \n\u003cli\u003efor_encoder_sun397.mat           -- serve as assist to your handle on sun397 dataset\u003c/li\u003e \n\u003cli\u003emit_pca.mat                      -- our generated scene_patchnet_feature pca matrix for mit indoor, used in vsad_encoding_example.m\u003c/li\u003e\n\u003cli\u003emit_vsad_codebook.mat            -- our generated semantical codebook for mit_indoor, used in vsad_encoding_example.m\u003c/li\u003e \n\u003cli\u003emulti_crop.m                     -- dense crop as 10 * 10 grid, used in extracting_feature_example.m\u003c/li\u003e \n\u003cli\u003eobject_selection_256.mat         -- 256 objects selected from 1000(in ImageNet), applied to both MIT_indoor and SUN397\u003c/li\u003e \n\u003cli\u003esun_pca.mat                      -- our generated scene_patchnet_feature pca matrix for sun397, used in vsad_encoding_example.m\u003c/li\u003e \n\u003cli\u003esun_vsad_codebook.mat            -- our generated semantical codebookfor sun397, used in vsad_encoding_example.m\u003c/li\u003e \n\u003cli\u003evsad_encoding_example.m          -- an example for VSAD encoding algorithm\u003c/li\u003e \n\u003cli\u003evsad_encoding.m                  -- our developed VSAD encoding function\u003c/li\u003e\n\u003cli\u003eplot_mit_sun.m                     -- Plot the figure in the below of this page\u003c/li\u003e\n\u003cli\u003exticklabel_rotate.m                  -- Serve for plot_mit_sun and rotate the text in the figure\u003c/li\u003e\n\u003c/ul\u003e\n\n\u003ch4\u003eUsage\u003c/h4\u003e \n\u003cp\u003e1. Download code and model\u003c/p\u003e\n\u003cp\u003e2. Extract scene_net_feature and object_net_probability (extracting_feature_example.m, multi_crop.m)\u003c/p\u003e\n\u003cp\u003e3. VSAD encoding (vsad_encoding.m, vsad_encoding_example.m, mit_pca.mat, mit_vsad_codebook.mat, object_selection_256.mat)\u003c/p\u003e\n\n\n\u003cp\u003eContact \u003c/p\u003e\n\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"http://wangzheallen.github.io/\"\u003eZhe Wang\u003c/a\u003e\u003c/li\u003e\n\u003cli\u003e\u003ca href=\"http://wanglimin.github.io/\"\u003eLimin Wang\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\n\u003cp\u003e Figure Plot for Reference \u003c/p\u003e\n\n![Alt text](https://github.com/wangzheallen/vsad/blob/master/4.png)\n\n\n\u003c/li\u003e\n\u003c/ul\u003e\u003c/h\u003e\u003c/td\u003e\n\u003c/tbody\u003e\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\u003c/table\u003e\u003c/p\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/strong\u003e\u003c/p\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fwangzheallen%2Fvsad","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fwangzheallen%2Fvsad","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fwangzheallen%2Fvsad/lists"}