{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibgcgl2vglfx5evvmynv7tdbcjtitfndmdrqilpetkigsrpb7mkma",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mks6rfmm44s2"
  },
  "path": "/t/real-time-exercise-form-analysis-with-mediapipe-looking-for-advice/175699#post_1",
  "publishedAt": "2026-05-01T11:40:20.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "Hey everyone,\n\nI’m a Computer and Communication Engineering student working on a course project with two teammates. We’re building a system called **FormAI** that uses a phone camera to analyze workout form in real time and give instant feedback .\n\nWe’re planning to use MediaPipe Pose for keypoint extraction, compute joint angles as features, and train classifiers to detect both the exercise type and whether the form is correct or not. For the dataset we’re going to use Fit3d (fit3d.imar.ro).\n\nWe have about a month and none of us have built a full computer vision pipeline before, so we’re looking for any advice, tips, or resources from people who’ve done something similar.\n\nAnything helps\nThanks a lot!",
  "title": "Real-time exercise form analysis with MediaPipe , looking for advice"
}