EfficientNet-B2 Food Classifier
Classifies food crops into 37 categories from the Nutrition5k dataset. Used as Stage 2 in a food segmentation + identification pipeline.
Stage 1 segmenter: arunapb/yolo11l-food-segmentation
Classes (37)
almond, apple, asparagus, avocado, broccoli, carrot, cauliflower, celery_stick, cherry, chicken_duck, corn, cucumber, dried_cranberries, egg, eggplant, enoki_mushroom, grape, green_beans, lettuce, melon, olives, onion, orange, pasta, pear, pepper, pineapple, pizza, potato, rice, salad, sausage, steak, tofu, tomato, watermelon, wonton_dumplings
Performance
Best validation accuracy: 0.9875
Usage
See class_names.json for the full class list. Load best_classifier.pth with torchvision EfficientNet-B2.