RectLabel is an offline image annotation tool designed for object detection and segmentation tasks. It offers a comprehensive suite of features to facilitate efficient and accurate labeling of images, catering to both individual users and larger deployments. Key Features and Functionality: - Advanced Annotation Tools: Supports labeling of polygons and pixels using models like Segment Anything Model 2 and Cellpose, as well as bounding boxes with tracking models. - Automatic Labeling: Utilizes Core ML models for automatic labeling and text recognition, enhancing productivity. - Versatile Labeling Options: Enables annotation of cubic bezier curves, line segments, points, oriented bounding boxes in aerial images, and keypoints with skeletons. - Pixel-Level Annotation: Provides tools for labeling pixels with brushes and superpixels for detailed segmentation. - Customizable Settings: Offers settings for objects, attributes, hotkeys, and fast labeling to tailor the tool to user preferences. - Efficient Management: Features a gallery view for searching objects, attributes, image names, and memos. - Flexible Export Options: Supports exporting to various formats, including COCO, Labelme, CreateML, YOLO, and DOTA, as well as indexed color and grayscale mask images. - Additional Utilities: Includes functionalities like converting videos to image frames and augmenting images. Primary Value and User Solutions: RectLabel streamlines the image annotation process, providing a robust and user-friendly platform for creating precise annotations essential for training machine learning models in object detection and segmentation. Its offline capability ensures data privacy and security, making it suitable for sensitive projects. The tool's versatility and comprehensive feature set address the diverse needs of users, from individual researchers to organizations, facilitating efficient and accurate image labeling workflows.