| model: | |
| framework: ultralytics-yolo | |
| task: pose-estimation | |
| variant: yolo11n-pose | |
| pretrained: true | |
| input_size: [640, 640] | |
| device: cpu | |
| precision: fp32 | |
| keypoints: | |
| schema: coco | |
| count: 17 | |
| flip_indices: | |
| - [1, 2] | |
| - [3, 4] | |
| - [5, 6] | |
| - [7, 8] | |
| - [9, 10] | |
| - [11, 12] | |
| - [13, 14] | |
| - [15, 16] | |
| training: | |
| epochs: 100 | |
| batch_size: 16 | |
| optimizer: auto | |
| initial_lr: 0.01 | |
| final_lr_factor: 0.01 | |
| momentum: 0.937 | |
| weight_decay: 0.0005 | |
| warmup_epochs: 3 | |
| loss_weights: | |
| box: 7.5 | |
| pose: 12.0 | |
| keypoint_objectness: 1.0 | |
| classification: 0.5 | |
| dfl: 1.5 | |
| data_augmentation: | |
| mosaic: 1.0 | |
| mixup: 0.0 | |
| cutmix: 0.0 | |
| horizontal_flip_prob: 0.5 | |
| vertical_flip_prob: 0.0 | |
| hsv: | |
| h: 0.015 | |
| s: 0.7 | |
| v: 0.4 | |
| scale: 0.5 | |
| translate: 0.1 | |
| auto_augment: randaugment | |
| erasing_prob: 0.4 | |
| deterministic: true | |
| seed: 0 | |
| evaluation: | |
| validation_split: val | |
| metrics: | |
| detection: | |
| precision: 0.995 | |
| recall: 0.952 | |
| mAP50: 0.979 | |
| mAP50-95: 0.938 | |
| pose: | |
| precision: 0.500 | |
| recall: 0.488 | |
| mAP50: 0.493 | |
| mAP50-95: 0.457 | |
| inference: | |
| confidence_threshold: 0.25 | |
| iou_threshold: 0.7 | |
| max_detections: 300 | |
| single_person_assumption: true | |
| temporal_smoothing: external | |
| optimized_for: | |
| - side-view footage | |
| - slow-motion video | |
| export: | |
| supported_formats: | |
| - torchscript | |
| - onnx | |
| simplify_onnx: true | |
| notes: | |
| use_case: sprint biomechanics and stride analysis | |
| camera_view: side | |
| video_type: slow-motion | |
| limitations: | |
| - Not optimized for frontal or oblique camera views | |
| - Performance may degrade with heavy occlusion | |
| - Single-athlete scenarios only |