1. Downloading the github repository

   git clone https://github.com/boostcampaitech3/final-project-level3-cv-16.git
   cd final-project-level3-cv-16/image_classification

2. Installing the requirements for training

   pip install -r requirements.txt
   pip install openpyxl

3. Download Pill Excel Data

   cd pill_excel_data

4. Preprocess Pill Data

   # follow the steps in /data_preprocessing

5. Train Pill Data for Image Classification

   python train.py

   usage: train.py [-h] [--epochs EPOCHS] [--batch_size BATCH_SIZE] [--patience PATIENCE] [--learning_rate LEARNING_RATE] [--lr_decay_step LR_DECAY_STEP]
                   [--accumulation_steps ACCUMULATION_STEPS] [--train_log_interval TRAIN_LOG_INTERVAL] [--seed SEED] [--opt OPT] [--sch SCH] [--excel_file_name EXCEL_FILE_NAME]
                   [--image_file_path IMAGE_FILE_PATH] [--project_type PROJECT_TYPE] [--user_name USER_NAME] [--model_name MODEL_NAME] [--project_name PROJECT_NAME]
                   [--train_whole TRAIN_WHOLE] [--create_test_data CREATE_TEST_DATA] [--test TEST] [--delete_pill_num DELETE_PILL_NUM [DELETE_PILL_NUM ...]] [--custom_label CUSTOM_LABEL]
   
   optional arguments:
     -h, --help            show this help message and exit
     --epochs EPOCHS       number of epochs to train (default: 50)
     --batch_size BATCH_SIZE
                           input batch size for training (default: 64)
     --patience PATIENCE   early stopping (default: 10)
     --learning_rate LEARNING_RATE
                           learning rate (defalt: 0.0001)
     --lr_decay_step LR_DECAY_STEP
                           learning rate deacy step (default: 5)
     --accumulation_steps ACCUMULATION_STEPS
                           training accumulation steps (default: 2)
     --train_log_interval TRAIN_LOG_INTERVAL
                           training log interval (default: 100)
     --seed SEED           fix seed (default: 16)
     --opt OPT             optimizer (default: Adam)
     --sch SCH             scheduler (default: StepLR)
     --excel_file_name EXCEL_FILE_NAME
                           name of the pill data excel (default: ./pill_excel_data/OpenData_PotOpenTabletIdntfc20220412.xls)
     --image_file_path IMAGE_FILE_PATH
                           path to image file (default: ./data/raw_data)
     --project_type PROJECT_TYPE
                           which column to use (default: shape)
     --user_name USER_NAME
                           user name (default: YH)
     --model_name MODEL_NAME
                           timm model name (default: resnet18)
     --project_name PROJECT_NAME
                           customize project name of what difference the project has (default: None)
     --train_whole TRAIN_WHOLE
                           train without validation set (default: False)
     --create_test_data CREATE_TEST_DATA
                           create test data from training dataset (default: False)
     --test TEST           do test with custom dataset (default: False)
     --delete_pill_num DELETE_PILL_NUM [DELETE_PILL_NUM ...]
                           list of file to delete (default: [])
     --custom_label CUSTOM_LABEL
                           customize labels for training (default: False)
   

   

6. pretrained model(resnet 50)

https://drive.google.com/file/d/1rDqbm3S9-0kWIhQngSuCM7VbLvmRUmbF/view?usp=sharing

# follow the steps in /kaggle_pill_data_preprocessing

Not used due to image normalization process done in Preprocess Pill Data

# follow the steps in /image_concatenation