This project implements U-Net, a convolutional neural network architecture, on the Oxford Pets III dataset. U-Net is particularly effective for image segmentation tasks, which makes it suitable for tasks like pet object detection, instance segmentation, and more.
The Oxford Pets III dataset contains a large collection of images of various pet breeds with corresponding pixel-wise masks for object segmentation. It serves as a benchmark for various computer vision tasks.
The dataset is downloaded from tensorflow datasets API inside the notebook and its size is approximately 770 mg.
To view and interact with the notebook containing the U-Net implementation and experimentations, please open the following file:
You can check out the model's architecture for u-net below this readme file, somehow it's available as an image file in this repository. (model.png)
- Clone the repository:
git clone https://github.com/parsakhavarinejad/unet_on_oxford_pet_dataset
- Install the required dependencies:
pip install -r requirements.txt
-
Open the notebook file in Jupyter Notebook or Jupyter Lab.
-
Follow the step-by-step instructions and code cells in the notebook to explore the U-Net implementation, train the model, evaluate results, and visualize predictions.
The model's accuracy is 81% in 15 epochs.
Contributions are welcome! If you find any issues or have suggestions for improvements, please create a GitHub issue or submit a pull request.