Training Pipeline
Click a node to filter hyperparameters
Label Studio
Source of annotated images with bounding boxes, tip / nock keypoints and scene landmarks.
get_stats()LS → YOLO Converter
Converts Label Studio JSON to YOLO-Pose .txt labels and per-image sidecar JSONs.
sync_annotations()aug_min_vis
Train / Val Split
Randomly splits image/label/sidecar triplets into train and validation sets.
_train_worker()val_splitseed
Data Augmentation
Pose-aware Albumentations pipeline — geometry, colour and noise transforms applied to training images only.
_build_pose_transform()aug_multiplieraug_resizedegreesfliplr+6
data.yaml
YOLO dataset config — paths, class names, kpt_shape=[2,3] (tip + nock), flip_idx.
training_service.pyimg_size
YOLO.train()
Trains YOLOv8-Pose on the prepared dataset. Runs in a background thread; progress is written to the database.
_train_worker()base_modelepochsbatch_sizeimg_size+6
trained_model/
Versioned best.pt weights saved to disk. Activate via the Run History page to swap the live model.
weights_path()Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.
Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.