Piston rod β€” experimental pose classifier

YOLOv8n-cls with three classes: correct_pose, flipped_pose, side_pose.

from ultralytics import YOLO
model = YOLO("weights/best.pt")
results = model.predict("crop.jpg")

Use part crops as input. Class IDs are in classes.json; the checkpoint records Ultralytics 8.4.51.

Validation top-1 Accuracy
Selected checkpoint 0.28523
Final logged epoch 0.24497

results.csv, metrics.json and training_config.json contain the training results and settings. Metrics describe validation performance.

Dataset: Pose classifier dataset. Paper and code: Vision assembly assistance. License: AGPL-3.0.

Downloads last month
21
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support