image-classification-level1-19

image-classification-level1-19 created by GitHub Classroom

P-Stage level-1 image classification competition. (shape of wearing mask, gender, age for a person)
19조 (BCAA) Solution

Archive contents

input/
├── data/
│ ├── eval
│ └── train
code/
├── train.py
├── inference.py
├── dataset.py
├── evaluation.py
├── loss.py
├── model.py
└── model
  └── exp1/

• input/data/eval: evaluation dataset
• input/data/train: train dataset
• code/train.py: main script to start training
• code/inference.py: evaluation of trained model
• code/dataset.py: custom data loader for dataset
• code/evaluation.py: function to score, matric
• code/loss.py: contains loss functions
• code/model.py: contains custom or pre-trained model
• model/: trained models are saved here like(exp1, exp2,...)
  

Requirements

• Linux version 4.4.0-59-generic
• Python >= 3.8.5
• PyTorch >= 1.7.1
• conda >= 4.9.2
• tensorboard >= 2.4.1

pip install -r requirements.txt : install the necessary packages.

Hadware

• CPU: Intel(R) Xeon(R) Gold 5220 CPU @ 2.20GHz
• GPU: Tesla V100-SXM2-32GB

Training

• general training with default args python train.py
• K-Fold training with default args python train.py --k_fold
• singel model training python train.py --single
• multiple model training python train.py --single --k_fold

Inference

python inference.py --model_dir {'MODEL_PATH'}
ex.
python inference.py --model_dir "./models/exp1"