I have a collection of images that must be precisely annotated with bounding boxes around three object classes: people, vehicles, and animals. Accuracy and consistency are critical because these annotations will feed directly into a computer-vision model. You will draw tight, non-overlapping boxes for every visible instance of the specified classes and save the labels in a standard format (COCO JSON, Pascal VOC XML, or YOLO TXT—let me know which of these you prefer). Tools such as CVAT, Labeling, or any comparable annotation software are fine as long as the output meets the chosen format’s schema. Before we begin, I will provide a short style guide that defines class names, edge-case handling, and file-naming conventions. Feel free to suggest improvements if you see ways to streamline the process. Deliverables: • Annotated files for all images in the agreed-upon format • A brief log summarizing any ambiguous cases or skipped frames • Confirmation that every file loads without errors in the target framework If you already have experience creating bounding boxes for large image datasets and can commit to quality control (spot checks or double-pass review), let’s get started right away.