DATA ANNOTATION/LABELLING
he project involved annotating image and video datasets to support the training of computer vision models. The primary objective was to accurately label objects, human features, and activities within video frames using polygon, point, and keypoint annotation techniques. Polygon annotation was used to outline objects with irregular shapes, ensuring precise boundary detection for model accuracy. Point annotation was applied for object localisation and counting tasks where detailed shape information was not required. Keypoint annotation was used to mark specific landmarks, such as human body joints, to enable pose estimation and movement analysis across video sequences. The work was carried out using professional annotation tools such as CVAT / Labelbox / Supervisely, following strict quality guidelines to ensure consistency, accuracy, and temporal continuity across frames. Each labelled dataset was reviewed to meet project standards and reduce annotation errors before submission.