lesson_49_synthetic_human_pose_estimation

Lesson 49: Synthetic Human Pose Estimation

This toy lesson renders simple grayscale stick figures and regresses eight normalized human keypoints:

• head
• left shoulder
• right shoulder
• left hand
• right hand
• pelvis
• left foot
• right foot

It follows the same small, CPU-friendly regression pattern as the nearby synthetic face landmark and face pose lessons.

Run a quick smoke test:

python -m tracks.vision.lesson_49_synthetic_human_pose_estimation.train \
  --epochs 1 \
  --max-train-batches 2 \
  --max-eval-batches 1 \
  --device cpu

Outputs land under outputs/vision/lesson_49_synthetic_human_pose_estimation/<run_name>/.