RM2025_ArmorDetect

# I Hiter Armor Plate Detection Dataset (2025) ## Dataset Introduction The **I Hiter Armor Plate Detection Dataset (2025)** is a dataset for robotic armor plate detection and pose estimation. It was collected and annotated by the **I Hiter Team during the 2025 RoboMaster season**. The dataset supports research in: - robotic perception - armor plate detection - keypoint-based pose estimation - real-time vision systems for robotics competitions Data sources include recordings from: - 2025 RoboMaster simulated competitions - exchange matches - regional competitions - national finals - historical datasets from previous seasons All annotations follow the **YOLO Pose format**, allowing direct use with modern detection frameworks. --- ## Dataset Statistics - **Total Images:** 21,156 - **Total Labels:** 21,156 - **Data Folders:** 78 - **Annotation Pairing Rate:** 100% --- ## Annotation Format The dataset uses **YOLO Pose annotation format**. Each armor plate is annotated with **four corner keypoints**. Keypoints are labeled in the following order: ``` top-left → bottom-left → bottom-right → top-right ``` Coordinates are normalized to the range **[0, 1]** relative to the image size. Example structure: ``` class_id x_center y_center width height kpt1_x kpt1_y kpt2_x kpt2_y kpt3_x kpt3_y kpt4_x kpt4_y ``` --- ## Usage The dataset can be used for training and evaluating: - armor plate detection models - keypoint-based pose estimation models - robotic targeting systems Supported frameworks include: - YOLOv8 - YOLO11 - YOLO26 --- ## License and Usage Restrictions This dataset is released **for research and educational purposes only**. Users must agree to the following conditions: 1. The dataset may only be used for academic or research purposes. 2. Commercial use is prohibited without permission. 3. Redistribution of the dataset is not allowed. 4. Proper citation is required when using this dataset. --- --- ## Contact For questions regarding the dataset: [e-mail] ---