Setup Env(MASK R-CNN)

Setup environment of mask R-CNN Run using Tensorflow 2.0.0 on Windowns using Docker and WSL¶

This tutorial walks through setting up Mask R-CNN using TensorFlow 2.0 on a Windows system with Docker and Windows Subsystem for Linux (WSL). By leveraging Docker, you avoid complex CUDA installations and ensure portability across systems.

Step 1: Activate WSL and Install Ubuntu¶

• There are two ways to activate WSL and install Ubuntu:

Option 1: Enable WSL via PowerShell¶

1. Open PowerShell as Administrator.

2. Run the following command:

   wsl --install
   

   

3. When prompted, choose Ubuntu as your Linux distribution.

   

Option 2: Install Ubuntu from Microsoft Store¶

1. Open Microsoft Store on Windows.

2. Search for Ubuntu and install the latest version.

   

3. Launch Ubuntu and complete the initial setup.

Step 2: Install Docker and Set Up TensorFlow Container¶

1. Install Docker Desktop from Docker's official website.

2. Open Ubuntu and run the following command to pull the TensorFlow Docker image with GPU support:

   

3. Verify that the image is available by checking Docker Desktop under Images.

   

Step 3: Set Up VS Code with Docker & Dev Containers¶

1. Open Visual Studio Code.

2. Install the following extensions (Ctrl+Shift+X → Search → Install):

   • Dev Containers

   • Docker

   

3. Attach VS Code to the TensorFlow container:

   • Press F1, type Remote-Containers: Attach to Running Container, and select the TensorFlow container.

   

Step 4: Load the Kangaroo Transfer Learning Project¶

• In VS Code’s terminal, add the Kangaroo Transfer Learning folder to your workspace:

  

Step 5: Update & Install Dependencies¶

1. Open the requirements.txt file in the project directory.

2. Modify package versions if needed.

   

3. Install dependencies inside the Docker container:

   pip install -r requirements.txt
   

   

Step 6: Download Pre-Trained Weights¶

1. Download the mask_rcnn_coco.h5 file from Mask R-CNN 2.0.

   • Download Link

2. Place the file in the appropriate project folder.

Step 7: Run the Kangaroo Training Script¶

1. Navigate to the project directory:

   cd kangaroo_transfer_learning
   

2. Start training:

   python3 kangaroo_training.py
   

Step 8: Test Model with a Sample Image¶

1. Locate an image in /kangaroo/images.

2. Rename the selected image to sample2.jpg.

3. Run the prediction script to evaluate the model:

   python3 kangaroo_prediction.py
   

Troubleshooting Common Issues¶

• ❌ ImportError: libGL.so.1 Not Found

• Fix by running:

  pip uninstall opencv-python
  pip install opencv-python-headless
  

Conclusion¶

By using Docker for machine learning, you achieve a reproducible, scalable, and efficient workflow for deep learning applications. With WSL and VS Code integration, developers can seamlessly train models while maintaining the flexibility of Linux and Windows environments.

🚀 Happy coding! 🖥️🐍📦