arunapb/yolo11l-food-segmentation

YOLO11l Food Segmentation — FoodSeg103 (103 classes)

Fine-tuned YOLO11l-seg for food ingredient instance segmentation on all 103 classes from the FoodSeg103 benchmark.

Model Details

Property	Value
Base model	yolo11l-seg (ImageNet pretrained)
Dataset	FoodSeg103 (103 food ingredient classes)
Image size	640 × 640
Epochs	60
Batch size	8
Task	Instance segmentation

Usage

from ultralytics import YOLO

model = YOLO("best.pt")
results = model.predict("food_image.jpg", conf=0.25)
results[0].show()

Classes (103)

French beans, almond, apple, apricot, asparagus, avocado, bamboo shoots, banana, bean sprouts, biscuit, blueberry, bread, broccoli, cabbage, cake, candy, carrot, cashew, cauliflower, celery stick, cheese butter, cherry, chicken duck, chocolate, cilantro mint, coffee, corn, crab, cucumber, date, dried cranberries, egg, egg tart, eggplant, enoki mushroom, fig, fish, french fries, fried meat, garlic, ginger, grape, green beans, hamburg, hanamaki baozi, ice cream, juice, kelp, king oyster mushroom, kiwi, lamb, lemon, lettuce, mango, melon, milk, milkshake, noodles, okra, olives, onion, orange, other ingredients, oyster mushroom, pasta, peach, peanut, pear, pepper, pie, pineapple, pizza, popcorn, pork, potato, pudding, pumpkin, rape, raspberry, red beans, rice, salad, sauce, sausage, seaweed, shellfish, shiitake, shrimp, snow peas, soup, soy, spring onion, steak, strawberry, tea, tofu, tomato, walnut, watermelon, white button mushroom, white radish, wine, wonton dumplings

Training Config

Parameter	Value
hsv_h	0.02
hsv_s	0.5
hsv_v	0.4
degrees	15
mosaic	1.0
mixup	0.1
close_mosaic	10
patience	15

Citation

@inproceedings{wu2021foodseg103,
  title     = {A Large-Scale Benchmark for Food Image Segmentation},
  author    = {Xiongwei Wu and Xin Fu and Ying Liu and Ee-Peng Lim and
               Steven C.H. Hoi and Qianru Sun},
  booktitle = {Proceedings of the 29th ACM International Conference on Multimedia},
  year      = {2021},
  doi       = {10.1145/3474085.3475628}
}