---
license: mit
language: en
library_name: pytorch
tags:
- image-inpainting
- computer-vision
- pytorch
- unet
- cifar-10
datasets:
- cifar10
---

# U-Net for Image Inpainting on CIFAR-10

This repository contains a PyTorch implementation of a deep U-Net with Residual Blocks, trained to perform image inpainting on the **CIFAR-10** dataset. The model takes a 32x32 image with a masked (blacked-out) region and reconstructs the missing part.

| Original | Masked Input | Reconstructed Output |
| :------: | :----------: | :------------------: |
| <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/diffusers-this-is-fine/original.png" width="128"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/diffusers-this-is-fine/masked.png" width="128"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/diffusers-this-is-fine/inpainted.png" width="128"> |
> **Note**: The images above are illustrative examples. You can generate your own by running the code below.

---

## Model Architecture

The model is a `ComplexUNet`, a variant of the standard U-Net architecture, designed to be deeper and wider for improved performance.

* **Framework**: PyTorch
* **Architecture**: U-Net with 4 downsampling and 4 upsampling stages.
* **Backbone**: Each stage uses **Residual Blocks** instead of simple convolutional layers.
* **Model Width**: The number of base channels is increased to `96` for higher capacity.
* **Total Parameters**: 73,148,259

---

## How to Use

The following code snippet provides a complete example of how to load the model, process an image from the CIFAR-10 test set, and visualize the inpainting result.

First, ensure you have the necessary libraries installed:
```bash
pip install torch torchvision numpy matplotlib Pillow