logeshsenthil13/Yolo26_Pose_2

YOLO26-Pose Benchmark (Small / Medium / Large)

Benchmarks the Ultralytics YOLO26-pose checkpoints from https://huggingface.co/Ultralytics/YOLO26 (yolo26s-pose.pt, yolo26m-pose.pt, yolo26l-pose.pt) on COCO Keypoints 2017 val across PyTorch FP32/FP16 -> ONNX FP32/FP16 -> ModelOpt INT8/INT4 -> TensorRT engines.

Yolo26_Pose/
  Dataset/download_coco_pose.sh   # builds Dataset/coco-pose (val2017 images, labels, keypoint GT)
  Small/  Medium/  Large/         # identical pipelines; only MODEL_NAME differs (see <Variant>/README.md)
  run_all_variants.sh             # runs all three sequentially, tee'ing <Variant>/full_run.log
  summarize_all.py                # combined results table

source /data/users/logesh/QC_AI_HUB/venv/bin/activate
bash Dataset/download_coco_pose.sh      # once
./run_all_variants.sh                   # everything (or ./run_all_variants.sh 200 for a smoke run)
python summarize_all.py                 # combined table any time

Or per variant: cd Small && ./run_all.sh (steps 01_* … 09_* can also be run one at a time; see Small/README.md).