{"id":21191808,"url":"https://github.com/edgeimpulse/pose-estimation-processing-block","last_synced_at":"2025-07-01T15:06:05.018Z","repository":{"id":104673443,"uuid":"472777688","full_name":"edgeimpulse/pose-estimation-processing-block","owner":"edgeimpulse","description":null,"archived":false,"fork":false,"pushed_at":"2025-03-05T08:44:57.000Z","size":4604,"stargazers_count":0,"open_issues_count":0,"forks_count":1,"subscribers_count":11,"default_branch":"master","last_synced_at":"2025-03-05T09:36:39.249Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause-clear","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/edgeimpulse.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":"2022-03-22T13:22:52.000Z","updated_at":"2025-03-05T08:45:00.000Z","dependencies_parsed_at":"2025-03-05T09:40:07.900Z","dependency_job_id":null,"html_url":"https://github.com/edgeimpulse/pose-estimation-processing-block","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/edgeimpulse%2Fpose-estimation-processing-block","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/edgeimpulse%2Fpose-estimation-processing-block/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/edgeimpulse%2Fpose-estimation-processing-block/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/edgeimpulse%2Fpose-estimation-processing-block/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/edgeimpulse","download_url":"https://codeload.github.com/edgeimpulse/pose-estimation-processing-block/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":243646664,"owners_count":20324586,"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":[],"created_at":"2024-11-20T19:05:22.083Z","updated_at":"2025-03-14T21:11:52.513Z","avatar_url":"https://github.com/edgeimpulse.png","language":"Python","readme":"# Pose estimation processing block\n\nThis implements a pose estimation [processing block](https://docs.edgeimpulse.com/docs/custom-blocks) based on [PoseNet](https://github.com/tensorflow/tfjs-models/tree/master/pose-detection) in Edge Impulse. Use this block to turn raw images into pose vectors, then pair it with an ML block to detect what a person is doing.\n\n## How to run this block (locally)\n\n1. Docker:\n    1. Build the container:\n\n        ```\n        $ docker build -t edge-detection-block .\n        ```\n\n    1. Run the block:\n\n        ```\n        $ docker run -p 4446:4446 -it --rm  edge-detection-block\n        ```\n\n1. Install [ngrok](https://ngrok.com) and open up port 4446 to the world:\n\n    ```\n    $ ngrok http 4446\n    ```\n\n    Note down the 'Forwarding' address that starts with https, e.g.:\n\n    ```\n    Forwarding                    https://4e9e1e61e3aa.ngrok.io -\u003e http://localhost:4446\n    ```\n\n1. In Edge Impulse, go to **Create impulse**, then:\n    1. Set the image width / height to 192 x 192 (this is the only resolution that works).\n    1. click *Add a processing block*, click *Add custom block* and enter the URL from the previous step.\n    1. Click *Add a learning block*, click *Classification*.\n    1. Click *Save impulse*.\n1. You now have pose estimation as a preprocessing block:\n\n    ![Pose estimation](images/pose-estimation.png)\n\n1. Train your model as usual 🚀\n\n## How to run this block (hosted in Edge Impulse)\n\nNote: this flow is only available for enterprise customers.\n\n1. Init the DSP block via:\n\n    ```\n    $ edge-impulse-blocks init\n    ```\n\n1. Push the block via:\n\n    ```\n    $ edge-impulse-blocks push\n    ```\n\n1. In Edge Impulse open a project owned by your organization, go to **Create impulse**, then:\n    1. Set the image width / height to 192 x 192 (this is the only resolution that works).\n    1. click *Add a processing block*, select the block.\n    1. Click *Add a learning block*, click *Classification*.\n    1. Click *Save impulse*.\n\n1. Follow the steps above!\n\n## Running on device\n\nThis block will run on Linux devices as-is. Just deploy as usual from the Studio.\n\n![inference demo](images/inference.jpg)\n\n### Updating the model\n\nDue to the size of the model and some unsupported ops it won't work on MCU in its current form. If you decide to train a smaller custom model, you'll need to replace `model.tflite`. You'll get feedback on the model through the 'On-device performance widget' in the Studio:\n\n\u003cimg src=\"images/ondevice-perf.png\" width=\"659\" title=\"On-device performance\"\u003e\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fedgeimpulse%2Fpose-estimation-processing-block","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fedgeimpulse%2Fpose-estimation-processing-block","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fedgeimpulse%2Fpose-estimation-processing-block/lists"}