Dessert Tray Segmentation Masks

I have a collection of food-photography images featuring assorted desserts presented on trays. To train an in-house vision model I need pixel-accurate segmentation masks that clearly separate every individual dessert from the tray it sits on. The job is strictly image annotation—no text, audio, or bounding-box work is involved. Each mask must outline the exact shape of the dessert or the tray (two distinct classes) so the background is fully excluded. Consistency across the whole set is essential because the dataset will feed directly into a semantic segmentation pipeline. Preferred tool is Roboflow that can export yolov8 format. Deliverables: • A folder of segmentation mask files (binary or indexed colour) that aligns one-to-one with the original image filenames. Acceptance criteria will be a visual spot-check on 10% of the images; masks must hug object edges without jagged gaps or over-spill, and class labels must be error-free. Once that sampling passes, the full milestone is approved. Let me know the approximate throughput you can maintain per day. Pricing can be per image.