[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"workflow-asset-bca17f13":3,"seo:featured-workflow:bca17f13-4ddd-11f1-9bc6-00163e2b0d79:fr":87,"workflow-related-asset-bca17f13-bca17f13-4ddd-11f1-9bc6-00163e2b0d79":88},{"id":4,"uuid":5,"slug":6,"title":7,"description":8,"author_id":9,"author_name":10,"author_avatar":11,"token_estimate":12,"time_saved":12,"model_used":11,"fork_count":12,"vote_count":12,"view_count":13,"parent_id":12,"parent_uuid":11,"lang_type":14,"steps":15,"tags":22,"has_voted":28,"visibility":18,"share_token":11,"is_featured":12,"content_hash":29,"asset_kind":30,"target_tools":31,"install_mode":35,"entrypoint":19,"risk_profile":36,"dependencies":38,"verification":46,"agent_metadata":49,"agent_fit":62,"trust":74,"provenance":83,"created_at":85,"updated_at":86},3249,"bca17f13-4ddd-11f1-9bc6-00163e2b0d79","asset-bca17f13","MMAction2 — OpenMMLab Video Understanding Toolbox","MMAction2 provides a modular framework for action recognition, temporal action detection, and spatial-temporal action detection with 20+ methods and support for major video benchmarks.","8a910e34-3180-11f1-9bc6-00163e2b0d79","Script Depot","",0,2,"en",[16],{"id":17,"step_order":18,"title":19,"description":11,"prompt_template":20,"variables":11,"depends_on":21,"expected_output":11},3812,1,"MMAction2 Video AI","# MMAction2 — OpenMMLab Video Understanding Toolbox\n\n## Quick Use\n```bash\npip install mmaction2 mmengine mmcv\npython demo\u002Fdemo.py \n    configs\u002Frecognition\u002Ftsn\u002Ftsn_imagenet-pretrained-r50_8xb32-1x1x3-100e_kinetics400-rgb.py \n    https:\u002F\u002Fdownload.openmmlab.com\u002Fmmaction\u002Frecognition\u002Ftsn\u002Ftsn_r50_1x1x3_100e_kinetics400_rgb.pth \n    demo\u002Fdemo.mp4 tools\u002Fdata\u002Fkinetics\u002Flabel_map_k400.txt\n```\n\n## Introduction\nMMAction2 is the next-generation video understanding toolbox from OpenMMLab. It covers action recognition, temporal action localization, and spatial-temporal action detection, providing a consistent PyTorch-based framework for researchers and practitioners working with video data.\n\n## What MMAction2 Does\n- Classifies human actions in video clips using 20+ recognition models\n- Localizes action segments temporally within untrimmed videos\n- Detects actions in space and time with spatial-temporal models\n- Supports skeleton-based action recognition via PoseC3D\n- Benchmarks on Kinetics, Something-Something, AVA, and more\n\n## Architecture Overview\nMMAction2 uses MMEngine as its training backend with a registry pattern for models, datasets, and pipelines. Recognition models process fixed-length clips through backbones like ResNet3D, SlowFast, or Video Swin Transformer. Temporal detectors use proposal generation and classification stages. All components are configured via Python config files.\n\n## Self-Hosting & Configuration\n- Install mmaction2, mmengine, and mmcv via pip\n- Download pre-trained checkpoints from the model zoo\n- Prepare video datasets in the expected directory structure\n- Modify config files for custom class labels and data paths\n- Use torchrun for multi-GPU distributed training\n\n## Key Features\n- Comprehensive coverage of action recognition paradigms (RGB, flow, skeleton)\n- UniFormerV2 and VideoMAE models achieve state-of-the-art on Kinetics\n- Modular design allows swapping backbones and temporal heads\n- Pre-built data pipelines for common video dataset formats\n- Integration with MMDeploy for production model conversion\n\n## Comparison with Similar Tools\n- **SlowFast (FAIR)** — reference implementation of the SlowFast network; MMAction2 includes SlowFast plus many other methods\n- **PyTorchVideo** — provides video-specific transforms and models; MMAction2 offers a broader set of methods and benchmarks\n- **TimeSformer** — single Transformer architecture; MMAction2 supports TimeSformer alongside CNN and hybrid approaches\n- **Decord** — video decoding library; MMAction2 uses Decord internally but adds full training and evaluation pipelines\n\n## FAQ\n**Q: Can I use MMAction2 for real-time action detection?**\nA: Yes. Lightweight models like MobileNetV2-TSM can run in real time on modern GPUs.\n\n**Q: Does it support skeleton-based recognition?**\nA: Yes. PoseC3D and ST-GCN models accept skeleton sequences extracted with MMPose.\n\n**Q: What video formats are supported?**\nA: MMAction2 reads any format supported by Decord or OpenCV, including MP4, AVI, and MKV.\n\n**Q: Can I fine-tune on my own action classes?**\nA: Yes. Update the label map and annotation files, then fine-tune from a Kinetics-pretrained checkpoint.\n\n## Sources\n- https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmaction2\n- https:\u002F\u002Fmmaction2.readthedocs.io\u002F","0",[23],{"id":24,"name":25,"slug":26,"icon":27},11,"Scripts","script","📜",false,"fc52b93e757d642599879a0899edd87d7519346d31a56efd74b073f064a5b9c8","skill",[32,33,34],"claude_code","codex","gemini_cli","single",{"executes_code":28,"modifies_global_config":28,"requires_secrets":37,"uses_absolute_paths":28,"network_access":28},[],{"npm":39,"pip":40,"brew":44,"system":45},[],[41,42,43],"mmaction2","mmcv","mmengine",[],[],{"commands":47,"expected_files":48},[],[19],{"asset_kind":30,"target_tools":50,"install_mode":35,"entrypoint":19,"risk_profile":51,"dependencies":53,"content_hash":29,"verification":58,"inferred":61},[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":52,"uses_absolute_paths":28,"network_access":28},[],{"npm":54,"pip":55,"brew":56,"system":57},[],[41,42,43],[],[],{"commands":59,"expected_files":60},[],[19],true,{"target":33,"score":63,"status":64,"policy":65,"why":66,"asset_kind":30,"install_mode":35},98,"native","allow",[67,68,69,70,71,72,73],"target_tools includes codex","asset_kind skill","install_mode single","markdown-only","policy allow","safe markdown-only Codex install","trust established",{"author_trust_level":75,"verified_publisher":28,"asset_signed_hash":29,"signature_status":76,"install_count":12,"report_count":12,"dangerous_capability_badges":77,"review_status":78,"signals":79},"established","hash_only",[],"unreviewed",[80,81,82],"author has published assets","content hash available","no dangerous capability badges",{"owner_uuid":9,"owner_name":10,"source_url":84,"content_hash":29,"visibility":18,"created_at":85,"updated_at":86},"https:\u002F\u002Ftokrepo.com\u002Fen\u002Fworkflows\u002Fasset-bca17f13","2026-05-12 16:36:57","2026-05-12 19:50:36",null,[89,140,187,238],{"id":90,"uuid":91,"slug":92,"title":93,"description":94,"author_id":9,"author_name":10,"author_avatar":11,"token_estimate":12,"time_saved":12,"model_used":11,"fork_count":12,"vote_count":12,"view_count":18,"parent_id":12,"parent_uuid":11,"lang_type":14,"steps":95,"tags":96,"has_voted":28,"visibility":18,"share_token":11,"is_featured":12,"content_hash":98,"asset_kind":30,"target_tools":99,"install_mode":35,"entrypoint":100,"risk_profile":101,"dependencies":103,"verification":108,"agent_metadata":111,"agent_fit":123,"trust":125,"provenance":128,"created_at":130,"updated_at":131,"__relatedScore":132,"__relatedReasons":133,"__sharedTags":138},3245,"7c63094f-4ddd-11f1-9bc6-00163e2b0d79","asset-7c63094f","MMSegmentation — OpenMMLab Semantic Segmentation Toolbox","MMSegmentation is a comprehensive semantic segmentation toolbox based on PyTorch, providing 50+ segmentation methods with a unified benchmark and modular design.",[],[97],{"id":24,"name":25,"slug":26,"icon":27},"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",[32,33,34],"SKILL.md",{"executes_code":28,"modifies_global_config":28,"requires_secrets":102,"uses_absolute_paths":28,"network_access":28},[],{"npm":104,"pip":105,"brew":106,"system":107},[],[],[],[],{"commands":109,"expected_files":110},[],[],{"asset_kind":30,"target_tools":112,"install_mode":35,"entrypoint":100,"risk_profile":113,"dependencies":115,"content_hash":98,"verification":120,"inferred":61},[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":114,"uses_absolute_paths":28,"network_access":28},[],{"npm":116,"pip":117,"brew":118,"system":119},[],[],[],[],{"commands":121,"expected_files":122},[],[],{"target":33,"score":63,"status":64,"policy":65,"why":124,"asset_kind":30,"install_mode":35},[67,68,69,70,71,72,73],{"author_trust_level":75,"verified_publisher":28,"asset_signed_hash":98,"signature_status":76,"install_count":12,"report_count":12,"dangerous_capability_badges":126,"review_status":78,"signals":127},[],[80,81,82],{"owner_uuid":9,"owner_name":10,"source_url":129,"content_hash":98,"visibility":18,"created_at":130,"updated_at":131},"https:\u002F\u002Ftokrepo.com\u002Fen\u002Fworkflows\u002Fasset-7c63094f","2026-05-12 16:35:09","2026-05-12 19:51:29",100.45154499349597,[134,135,136,137],"topic-match","same-kind","same-target","same-author",[26,139],"scripts",{"id":141,"uuid":142,"slug":143,"title":144,"description":145,"author_id":9,"author_name":10,"author_avatar":11,"token_estimate":12,"time_saved":12,"model_used":11,"fork_count":12,"vote_count":12,"view_count":146,"parent_id":12,"parent_uuid":11,"lang_type":14,"steps":147,"tags":148,"has_voted":28,"visibility":18,"share_token":11,"is_featured":12,"content_hash":150,"asset_kind":30,"target_tools":151,"install_mode":35,"entrypoint":152,"risk_profile":153,"dependencies":155,"verification":160,"agent_metadata":163,"agent_fit":175,"trust":177,"provenance":180,"created_at":182,"updated_at":183,"__relatedScore":184,"__relatedReasons":185,"__sharedTags":186},2458,"7e2317bb-453a-11f1-9bc6-00163e2b0d79","cogvideo-text-image-video-generation-7e2317bb","CogVideo — Text and Image to Video Generation","An open-source video generation framework from Zhipu AI supporting text-to-video and image-to-video with CogVideoX models. Generates high-quality clips up to 6 seconds.",81,[],[149],{"id":24,"name":25,"slug":26,"icon":27},"4a26ff53462e623ca42b9cec2ef2df5a01afa8fe0735e8fdf42b5dad217a44a6",[32,33,34],"CogVideo Overview",{"executes_code":28,"modifies_global_config":28,"requires_secrets":154,"uses_absolute_paths":28,"network_access":28},[],{"npm":156,"pip":157,"brew":158,"system":159},[],[],[],[],{"commands":161,"expected_files":162},[],[152],{"asset_kind":30,"target_tools":164,"install_mode":35,"entrypoint":152,"risk_profile":165,"dependencies":167,"content_hash":150,"verification":172},[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":166,"uses_absolute_paths":28,"network_access":28},[],{"npm":168,"pip":169,"brew":170,"system":171},[],[],[],[],{"commands":173,"expected_files":174},[],[152],{"target":33,"score":63,"status":64,"policy":65,"why":176,"asset_kind":30,"install_mode":35},[67,68,69,70,71,72,73],{"author_trust_level":75,"verified_publisher":28,"asset_signed_hash":150,"signature_status":76,"install_count":12,"report_count":12,"dangerous_capability_badges":178,"review_status":78,"signals":179},[],[80,81,82],{"owner_uuid":9,"owner_name":10,"source_url":181,"content_hash":150,"visibility":18,"created_at":182,"updated_at":183},"https:\u002F\u002Ftokrepo.com\u002Fen\u002Fworkflows\u002Fcogvideo-text-image-video-generation-7e2317bb","2026-05-01 16:48:15","2026-05-12 20:11:21",85.87072077857557,[134,135,136,137],[26,139],{"id":188,"uuid":189,"slug":190,"title":191,"description":192,"author_id":193,"author_name":194,"author_avatar":11,"token_estimate":12,"time_saved":12,"model_used":11,"fork_count":12,"vote_count":12,"view_count":195,"parent_id":12,"parent_uuid":11,"lang_type":14,"steps":196,"tags":197,"has_voted":28,"visibility":18,"share_token":11,"is_featured":12,"content_hash":98,"asset_kind":30,"target_tools":203,"install_mode":35,"entrypoint":100,"risk_profile":204,"dependencies":206,"verification":211,"agent_metadata":214,"agent_fit":226,"trust":228,"provenance":231,"created_at":233,"updated_at":234,"__relatedScore":235,"__relatedReasons":236,"__sharedTags":237},3246,"8d12bfc9-4ddd-11f1-9bc6-00163e2b0d79","asset-8d12bfc9","MMPose — OpenMMLab Pose Estimation Toolbox","MMPose provides a modular framework for 2D and 3D pose estimation covering human body, hand, face, and animal keypoint detection with 30+ state-of-the-art methods.","8a911193-3180-11f1-9bc6-00163e2b0d79","AI Open Source",4,[],[198],{"id":199,"name":200,"slug":201,"icon":202},12,"Configs","config","⚙️",[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":205,"uses_absolute_paths":28,"network_access":28},[],{"npm":207,"pip":208,"brew":209,"system":210},[],[],[],[],{"commands":212,"expected_files":213},[],[],{"asset_kind":30,"target_tools":215,"install_mode":35,"entrypoint":100,"risk_profile":216,"dependencies":218,"content_hash":98,"verification":223,"inferred":61},[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":217,"uses_absolute_paths":28,"network_access":28},[],{"npm":219,"pip":220,"brew":221,"system":222},[],[],[],[],{"commands":224,"expected_files":225},[],[],{"target":33,"score":63,"status":64,"policy":65,"why":227,"asset_kind":30,"install_mode":35},[67,68,69,70,71,72,73],{"author_trust_level":75,"verified_publisher":28,"asset_signed_hash":98,"signature_status":76,"install_count":12,"report_count":12,"dangerous_capability_badges":229,"review_status":78,"signals":230},[],[80,81,82],{"owner_uuid":193,"owner_name":194,"source_url":232,"content_hash":98,"visibility":18,"created_at":233,"updated_at":234},"https:\u002F\u002Ftokrepo.com\u002Fen\u002Fworkflows\u002Fasset-8d12bfc9","2026-05-12 16:35:37","2026-05-12 20:40:50",84.04845500650403,[134,135,136],[],{"id":239,"uuid":240,"slug":241,"title":242,"description":243,"author_id":9,"author_name":10,"author_avatar":11,"token_estimate":244,"time_saved":12,"model_used":245,"fork_count":12,"vote_count":12,"view_count":246,"parent_id":12,"parent_uuid":11,"lang_type":14,"steps":247,"tags":248,"has_voted":28,"visibility":18,"share_token":11,"is_featured":12,"content_hash":250,"asset_kind":30,"target_tools":251,"install_mode":35,"entrypoint":242,"risk_profile":252,"dependencies":254,"verification":259,"agent_metadata":262,"agent_fit":274,"trust":276,"provenance":280,"created_at":282,"updated_at":283,"__relatedScore":284,"__relatedReasons":285,"__sharedTags":286},104,"57b68929-f919-474a-99f1-81ac76e0bce6","moviepy-python-video-editing-library-57b68929","MoviePy — Python Video Editing Library","Python library for video editing: cutting, concatenating, adding titles, effects, and audio. Script-based video production for automation pipelines.",500,"Claude Code",390,[],[249],{"id":24,"name":25,"slug":26,"icon":27},"e5e02dd5dbc90f55da884bf59e06291d7f4f1ef1282bb146817b5b827b310c22",[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":253,"uses_absolute_paths":28,"network_access":28},[],{"npm":255,"pip":256,"brew":257,"system":258},[],[],[],[],{"commands":260,"expected_files":261},[],[242],{"asset_kind":30,"target_tools":263,"install_mode":35,"entrypoint":242,"risk_profile":264,"dependencies":266,"content_hash":250,"verification":271},[32,33,34],{"executes_code":28,"modifies_global_config":28,"requires_secrets":265,"uses_absolute_paths":28,"network_access":28},[],{"npm":267,"pip":268,"brew":269,"system":270},[],[],[],[],{"commands":272,"expected_files":273},[],[242],{"target":33,"score":63,"status":64,"policy":65,"why":275,"asset_kind":30,"install_mode":35},[67,68,69,70,71,72,73],{"author_trust_level":75,"verified_publisher":28,"asset_signed_hash":250,"signature_status":76,"install_count":12,"report_count":12,"dangerous_capability_badges":277,"review_status":78,"signals":278},[],[279,80,81,82],"asset has usage views",{"owner_uuid":9,"owner_name":10,"source_url":281,"content_hash":250,"visibility":18,"created_at":282,"updated_at":283},"https:\u002F\u002Ftokrepo.com\u002Fen\u002Fworkflows\u002Fmoviepy-python-video-editing-library-57b68929","2026-03-29 16:58:49","2026-05-12 19:12:34",83.88826513609379,[134,135,136,137],[26,139]]