| --- |
| license: mit |
| base_model: |
| - AXERA-TECH/YOLO11-Pose |
| pipeline_tag: keypoint-detection |
| tags: |
| - sports |
| - swimming |
| - pose-estimation |
| --- |
| |
| # SwimAnalysisPro: YOLO11-Pose for Side-View Swimming Analysis |
| # 側面游泳姿勢分析模型 |
|
|
| ## Model Description / 模型描述 |
|
|
| This model is fine-tuned based on the Ultralytics YOLO11-pose architecture, specifically optimized for detecting and tracking swimmer poses from a side-view perspective. It is designed for technique analysis and biomechanical evaluation in aquatic environments. |
|
|
| 本模型基於 Ultralytics YOLO11-pose 架構進行微調，專門針對側面視角的游泳者姿勢進行偵測與追蹤優化。適用於游泳技術分析以及水下或水面的生物力學評估。 |
|
|
| ## Training Dataset / 訓練資料集 |
|
|
| The model was trained on a large-scale specialized dataset to ensure robustness across different styles. |
|
|
| 本模型使用大規模專用資料集進行訓練，確保在不同泳姿下的穩定性。 |
|
|
| - Total Images / 影像總數: 28,000 |
| - Perspective / 視角: Side-view / 側面視角 |
| - Stroke Coverage / 涵蓋泳姿: Butterfly, Backstroke, Breaststroke, and Freestyle / 蝶式、仰式、平式、自由式全涵蓋 |
|
|
| ## Keypoints Index / 關鍵點索引表 |
|
|
| The model tracks 7 core human keypoints optimized for swimming stroke analysis: |
|
|
| 本模型追蹤 7 個核心人體關鍵點，針對游泳動作分析進行優化： |
|
|
| | Index / 索引 | Keypoint (English) | 關鍵點 (中文) | |
| | --- | --- | --- | |
| | 0 | Head | 頭部 | |
| | 1 | Shoulder | 肩膀 | |
| | 2 | Elbow | 手肘 | |
| | 3 | Wrist | 手腕 | |
| | 4 | Hip | 髖部 | |
| | 5 | Knee | 膝蓋 | |
| | 6 | Ankle | 腳踝 | |
|
|
| ## Performance / 效能指標 |
|
|
| The evaluation metrics including mAP50 and mAP50-95 are provided in the charts below. |
|
|
| 測試集上的效能數據（如 mAP50 與 mAP50-95）請參考下方圖表。 |
|
|
|  |
|
|
| ## How to Use / 如何使用 |
|
|
| You can load and run this model directly using the Ultralytics Python package: |
|
|
| 你可以使用 Ultralytics Python 套件直接載入並執行此模型： |
|
|
| ```python |
| from ultralytics import YOLO |
| |
| # Load the model |
| model = YOLO("your_username/your_model_name") |
| |
| # Run inference |
| results = model.predict(source="swimming_video.mp4", save=True, conf=0.5) |
| |
| # Process results |
| for result in results: |
| keypoints = result.keypoints.data |
