FoodSeg103 Dataset is a dataset for semantic segmentation and object detection tasks. It is used in the food industry.
The dataset consists of 7118 images with 26016 labeled objects belonging to 103 different classes including bread, carrot, chicken duck, and other: sauce, tomato, potato, steak, broccoli, ice cream, cilantro mint, rice, pork, lemon, lettuce, strawberry, pie, cucumber, onion, corn, cake, pepper, cheese butter, french beans, fish, biscuit, egg, asparagus, noodles, and 75 more.
Images in the FoodSeg103 dataset have pixel-level semantic segmentation annotations. All images are labeled (i.e. with annotations). There are 2 splits in the dataset: train (4983 images) and test (2135 images). Alternatively, the dataset could be split into 15 supercategories: vegetable (9635 objects), main (3665 objects), meat (3561 objects), fruit (3007 objects), dessert (2429 objects), sauce (1145 objects), beverage (714 objects), seafood (560 objects), fungus (303 objects), nut (299 objects), egg (292 objects), other ingredients (227 objects), soup (89 objects), tofu (68 objects), and salad (22 objects). The dataset was released in 2021 by the Management University, Singapore and Beijing Jiaotong University, China.
Here is a visualized example for randomly selected sample classes: