Level1 P-stage Image classification

💡 Team: boos2

Project Overview

• Predict mask / gender / age
• Input : 384 x 512 Image
• Output : 0~17 classes
  • Mask : Wear, Incorrect, Not Wear
  • Gender : Male, Female
  • Age : <30, >=30 and <60, >=60

Archive contents

image-classification-level1-02/
├── input/
│   └── train/
|		├── images/
|		└── train.csv
│   └── eval/
|		├── images/
|		└── info.csv
├── Dataset.py
├── Model.py
├── Test.py
├── functions.py
└── main.py

• input/data/train : train dataset images
• input/data/test : used in evaluation
• input/ : download from https://stages.ai/

Requirements

• Ubuntu 18.04.5
• Python 3.8.5
• pytorch 1.7.1
• torchvision 0.8.2

Install packages : pip install -r requirements.txt

Hardware

• CPU: 8 x Intel(R) Xeon(R) Gold 5220 CPU
• GPU: V100
• RAM: 88GB

Train Model & Test & Save Result

sh run.sh

run.sh

python main.py \
--PATH '../input/data' \
--BATCH_SIZE 128 \
--SAVE true \
--SAVE_PATH 'saved/' \
--EPOCH 20

• --PATH: train/, eval/ parent directory
• --BATCH_SIZE: Batch size (default=128)
• --SAVE: If want to save weights while training else remove this line
• --SAVE_PATH: Saving weights directory
• --EPOCH: Train epoch (default=20)

Trained Model

• Pretrained ResNet18 model
• No data augmentation
• Hyperparameter
  • Learning rate : 1e-5
  • Optimizer : Adam
  • Batch size : 128

Contributors

Name @github	Project link
안명철 @abbymark	image classification
김민진 @kkmjkim	image classification
박상현 @hyun06000	image classification
심세령 @seryoungshim17	image classification
이유진 @YoojLee	image classification
정두해 @Doohae	image classification
홍지연 @hongjourney	image classification