How sarinkaavya19-ai/emo-affectnetmodel works
This repository appears to center on a webcam-based facial-emotion demo: capture frames, detect/track a face, run emotion inference, and display or record results. The dependency notes document webcam capture, MediaPipe FaceLandmarker, TorchScript inference, overlays, and VideoWriter output; the model and face-task assets are present in the tree. The notebook itself and architecture documents were not sampled, so finer control flow and any additional capabilities cannot be verified. The older pipeline under _unused is excluded.
Main entry points: check_backbone_models_by_webcam.output.ipynb Demo workflow · torchscript_model_0_66_49_wo_gl.pth Emotion model · face_landmarker.task Face task model
Components 10
Webcam demo
- Demo workflow
check_backbone_models_by_webcam.output.ipynb - Frame capture
check_backbone_models_by_webcam.output.ipynb - Face landmarks
check_backbone_models_by_webcam.output.ipynb - Emotion inference
check_backbone_models_by_webcam.output.ipynb
Model assets
- Emotion model
models_EmoAffectnet/torchscript_model_0_66_49_wo_gl.pth - Face task model
models_face/face_landmarker.task
Results
- Annotated preview
check_backbone_models_by_webcam.output.ipynb - Recorded video
Other components
- Operator
- Webcam
Connections 9
- Operator → Demo workflow: runs
- Webcam → Frame capture: supplies frames
- Demo workflow → Frame capture: uses
- Demo workflow → Face landmarks: uses
- Demo workflow → Emotion inference: uses
- Demo workflow → Annotated preview: renders
- Demo workflow → Recorded video: writes
- Face landmarks → Face task model: loads
- Emotion inference → Emotion model: loads