The repository contains the source code for training the model proposed in the paper for food image recognition tasks.

The source code can be obtained in the local python environment using the below command:

git clone [email protected]:LduIIPLab/CVnets.git
cd CVnets/FoodProject
pip install -r requirements.txt

python main_train.py

• You can directly run the above code on the server terminal to train the model.
• Before training, you need to modify the path of datasets in config files for different datasets.
• If you want to train different data or change the size of the model, you can achieve this by modifying the code below in the main_train.py file.
  • "food101": ETHZ Food-101
  • "food172": Vireo Food-172
  • "food256": UEC Food-256

experiments_config.ehfr_net_config_forward(dataset_name="food101", width_multiplier=0.5)

The images should be stored in the following two formats:

Food-101/
├── train
│   ├── 1
│   │   ├── 1_2.JPEG
│   │   ├── 1_3.JPEG
...
├── val
│   ├── 1
│   │   ├── 1_6.JPEG
...

• We provide a dataset called "food" to read the image in the storage format above and modify the relevant parts in the config file like the code provided below

dataset:
  root_train: "/ai35/Food/Food-101"
  root_val: "/ai35/Food/Food-101"
  train_index_file: "train_full.txt"
  val_index_file: "val_full.txt"
  name: "food"

Food-101/
├── apple_pie
│   ├── 134.JPEG
│   ├── 21063.JPEG
...
├── baby_back_ribs
│   ├── 2432.JPEG
...
├── train_full.txt
├── val_full.txt

• We provide another dataset called "food_annother" to read the image in the storage format above and modify the relevant parts in the config file like the code provided below

dataset:
  root_train: "/ai35/Food/Food-101"
  root_val: "/ai35/Food/Food-101"
  name: "food_another"

• If you want to train your own dataset, please put the format of the dataset into the folder according to the two formats provided above, and generate a new yaml file for configuring the config. The name format is "model name_dataset file name. yaml", and make the relevant configuration.
• In addition, it is necessary to add the code for training one's own model in the form of internal code in "experiments.config. py", and modify the code at the corresponding position in "main_train. py".

If you find our project is helpful, please feel free to leave a star and cite our paper:

@article{sheng2024EHFR-Net
	title={A Lightweight Hybrid Model with Location-Preserving ViT for Efficient Food Recognition},
	author={Guorui Sheng, Weiqing Min, Xiangyi Zhu, Liang Xu, Qingshuo Sun, Yancun Yang, Lili Wang and Shuqiang Jiang},
	journal={Nutrients},
	year={2024}
}

• License