Boostcamp AI-Tech 2nd - Master Slave Team

Level 1 P-Stage Image Classification Project Repository

Members

• 김대웅(팀장) : 스케쥴 및 할 일 관리, 전체 프레임워크 잡기
• 나요한 : Augmentation 실험, 기본 모델들의 성능 평가 및 모델 구현에 기여
• 박준수 : Augmentation 실험, 기본 모델들의 성능 평가 및 모델 구현에 기여
• 이호영 : 데이터 EDA, 추가 데이터셋 확보에 기여
• 최한준 : 데이터 EDA, 추가 데이터셋 확보에 기여
• 추창한 : 멀티 모델 실험, 모델의 성능 저하원인 분석
• 한진 : 멀티 모델 실험, 모델의 성능 저하원인 분석

Architecture

• Overview

• vit_base_patch16_384

  

• Efficientnet_b3

  

Requirements

• Python >= 3.5 (3.8 recommended)
• PyTorch >= 1.4 (1.9 recommended)
• tqdm (Optional for test.py)
• efficientnet_pytorch
• albumnetations
• torchmetrics
• timm
• wandb

Folder Structure

  input/data/
  ├──train
  │   ├── images
  │   │    ├ 000001_female_Asian_45
  │   │    │  ├ mask1.jpg
  │   │    │  ├ mask2.jpg
  │   │    │  ├ mask3.jpg
  │   │    │  ├ mask4.jpg
  │   │    │  ├ mask5.jpg
  │   │    │  ├ incorrect.jpg
  │   │    │  └ normal.jpg                          
  │   │    └ ...
  │   ├── train.csv
  │   └── trans_train.csv  
  └──eval
  ├── info.csv
  └── images
       ├── ._0a2bd33bf76d7426f3d6ca0b7fbe03ee431159b4.jpg
       └── ...
  image-classification-level1-04/
  │
  ├── train.py - main script to start training
  ├── test.py - evaluation of trained model
  │
  ├── config.json - holds configuration for training
  ├── config_submit.json - holds configuration for submit
  │
  ├── parse_config.py - class to handle config file and cli options
  │  
  ├── base/ - abstract base classes
  │   ├── base_data_loader.py
  │   ├── base_model.py
  │   └── base_trainer.py
  │
  ├── data_loader/ - anything about data loading goes here
  │   ├── data_loader.py
  │   ├── datasets.py
  │   └── transforms.py
  │
  ├── model/ - models, losses, and metrics
  │   ├── model.py
  │   ├── metric.py
  │   └── loss.py
  │
  ├── saved/
  │   ├── models/ - trained models are saved here
  │   └── log/ - default logdir for tensorboard and logging output
  │
  ├── trainer/ - trainers
  │   └── trainer.py
  │
  ├── logger/ - module for tensorboard visualization and logging
  │   ├── visualization.py
  │   ├── logger.py
  │   └── logger_config.json
  │  
  └── utils/ - small utility functions
  ├── util.py
  └── ...

How to use this Code

# BaseModel
python3 train.py -c config_main.json
# AgeModel
python3 train.py -c config_sub.json

# Submit
python3 submit.py -c config_submit_main.json
python3 submit.py -c config_submit_sub.json

Config file format

Config files are in .json format:

{
    "name": "PretrainModelTimmViTBasePath16_TRNS004_Adam_StepLR",  // training session name
    "n_gpu": 1,  // number of GPUs to use for training.
    "arch": {
        "type": "PretrainModelTimmViTBasePath16",                  // name of model architecture to train
        "args": {}
    },
    "transforms_select": {
        "type": "transforms_select",                               // selecting transforms methods
        "args": {
            "method": "VIT_TRNS004",
            "default": "VIT_DEFAULT"
        }
    },
    "data_loader": {
        "type": "MaskDataLoader",         			   // selecting data loader                
        "args": {
            "data_dir": "../input/data",  		    	   // dataset path
            "batch_size": 32,             			   // batch size
			"shuffle": true,                           // shuffle training data before splitting
			"validation_split": 0.1          	   // size of validation dataset. float(portion) or int(number of samples)
			"num_workers": 2,                	   // number of cpu processes to be used for data loading
			"trsfm": false,              		   // use transforms
			"submit": false                		   // submission

        }
    },
    "optimizer": {
        "type": "Adam",
        "args": {
            "lr": 2e-5,                     			   // learning rate
            "weight_decay": 0,                			   // (optional) weight decay
            "amsgrad": true
        }
    },
    "loss": {
        "type": "cross_entropy_loss",				   //loss
        "args": {
            "class_weight": false
        }
    },
    "metrics": [
        "accuracy","f1"						   // list of metrics to evaluate
    ],
    "lr_scheduler": {
        "type": "StepLR",					   // learning rate scheduler
        "args": {
            "step_size": 50,
            "gamma": 0.1
        }
    },
    "trainer": {
		"epochs": 20,                                      // number of training epochs
		"save_dir": "saved/",              		   // checkpoints are saved in save_dir/models/name
		"save_period": 1,                    		   // save checkpoints every save_freq epochs
		"verbosity": 2,                    		   // 0: quiet, 1: per epoch, 2: full	
		"monitor": "min val_loss"          		   // mode and metric for model performance monitoring. set 'off' to disable.
		"early_stop": 10,	                	   // number of epochs to wait before early stop. set 0 to disable.  		
    },
    "wandb": {
        "use": true,						   //enable tensorboard visualization
        "args": {
            "project": "basic", 				   //sub project name
            "entity": "boostcamp-level01-04"			   //project name
        }
    }
}

Using config files

Modify the configurations in .json config files, then run:

python train.py --config config.json

Resuming from checkpoints

You can resume from a previously saved checkpoint by:

python train.py --resume (e.g. saved/models/[confg.name]/[MMDD_Hashvalue]/checkpoint-epoch#.pth)

Test from checkpoints

You can test from a previously saved checkpoint by:

python test.py --resume path/to/checkpoint(e.g. saved/models/[confg.name]/[MMDD_Hashvalue]/checkpoint-epoch#.pth)

Submit from checkpoints

You can submit from a previously saved checkpoint by:

python submit.py --resume path/to/checkpoint(e.g. saved/models/[confg.name]/[MMDD_Hashvalue]/checkpoint-epoch#.pth)

TODOs

• Add transforms feature(or Albumentation)
• Weights & Biases logger support
• Add pretrained model(e.g. EfficientNet, ViT, Resnet ...)
• Data Augmentation (e.g. CLAHE, Elastic & Cutmix, Horizontal Flip, ...)
• Soft, Hard Ensemble
• Time Test Augment
• Multi Sample Dropout
• Focal Loss
• ArcFace Loss
• Label Smoothing Loss
• Angular Additive Margin Loss
• OverSampling
• K-Fold Validate
• Crop using MTCNN

License

This project is licensed under the MIT License. See LICENSE for more details

Acknowledgements

• We use this project template(PyTorch-template) by victoresque

• Imbalanced Dataset Sampler

• MaskTheFace

• All-Age-Faces-Dataset