DAC4TB

Training

Installation

1. Install CUDA Toolkit https://developer.nvidia.com/cuda-10.1-download-archive-update2
2. Install Anaconda https://www.anaconda.com/distribution/#download-section

Run Training

1. Open Jupyter Notebook from Anaconda Navigator
2. Open file predict.ipynb
3. Copy datasets into folder Dataset\DATASET_NAME
   Seperate each images labels into sub folder:
   Dataset\DATASET_NAME\ABNORMAL
Dataset\DATASET_NAME\NORMAL
   Only *.jpg is allow.
4. Change dataset name in line dataset_name = "DDC Prison BLM TUA"
5. Change model name in line MODEL_NAME = 'blm_2_t_mn'
   Training model will be save into folder saved_model/
6. Click run to training

Predict

Installation

1. Install docker https://www.docker.com/products/docker-desktop
2. Right click from docker icon in taskbar then click Settings
3. Click Resources\FILE SHARING then tick drive that you want to save input and output folder.
   Default is drive C:.
   Then click Apply & Restart Button
4. Increase docker cpus and memory resource from Resources\ADVANCED then click Apply & Restart Button
5. Create input and output folders in C:\input and C:\output

Run Prediction

1. Copy images that want to labels into c:\input\000xxxx.jpg.
   Only *.jpg is allow.
2. Open Command Prompt then run these command.
   docker run -v c:/input:/input -v c:/output:/output -it asia.gcr.io/thaihealthai/dac4tb:blm_2_t_mn-1.0.0 python predict.py
3. The results will be in folder c:\output\output_yyyymmmdddhhmmss.csv
4. If input and output folders different from default. Please change -v d:/input_folder:/input -v d:/output_folder/:output.
5. Change other dac4tb version from asia.gcr.io/thaihealthai/dac4tb:blm_2_t_mn-1.0.0
6. Easy way to predict is double click predict.cmd