📄 Paper | 💻 Source Code | 🤗 Hugging Face
Architecture of the CR-Net model.
## Introduction **CR-Net** is a model enhance the quality of images and videos captured under low-light conditions. By learning a continuous rendering process, CR-Net effectively improves brightness, producing natural and sharp results even in challenging dark environments. To learn more about CR-Net, feel free to read our documentation [English](../README.md) | [Tiếng Việt](preview/README-vi.md) | [ä¸æ–‡](preview/README-zh.md).
Smooth continuous light to dark transition with phi angle
### Key Features * **Low-light image/video enhancement:** Significantly improves brightness and contrast for images and videos captured in dim lighting. * **Continuous rendering network:** Employs a novel architecture to deliver smoother and more natural results compared to traditional methods. * **Flexible applications:** Supports both video processing and directories containing multiple still images. ## Demo  ## Installation and Requirements To run this model, you need the proper environment. We recommend the following versions: * **Python:** `Python >= 3.10` (Recommended `Python 3.10`) * **PyTorch:** `PyTorch >= 1.12` (Recommended `PyTorch 2.1.2`) **Step 1: Clone the repository** ```shell git clone https://github.com/val-utehy/CR-Net.git cd CR-Net ``` **Step 2: Install dependencies** ```shell pip install -r requirements.txt ``` > [!NOTE] > Make sure you have installed the compatible versions of **torch** and **torchvision** with your **CUDA driver** to leverage GPU. ## Pretrained Models You can download the pretrained models from this [link](https://huggingface.co/val-utehy/CR-Net/tree/main/checkpoints_v2/ast_rafael_v2_sharpening). You can use latest checkpoint `latest_net_G.pth` and `opt.pkl` for inference. > [!NOTE] > Please ensure your path to the checkpoint and config (opt.pkl) is correct in the script files before running. ## Usage Guide ### 1. Model Training Training file will be updated soon! [//]: # (To train the CR-Net model on your own dataset, follow these steps:) [//]: # () [//]: # (**a. Configure the training script file:**) [//]: # () [//]: # (Open and edit the file `train_scripts/ast_n2h.sh`. In this file, you need to specify important paths such as the dataset path and the checkpoint saving directory.) [//]: # () [//]: # (**b. Run the training script:**) [//]: # () [//]: # (After finishing the configuration, navigate to the project’s root directory and execute the following command:) [//]: # () [//]: # (```shell) [//]: # ( bash train_scripts/ast_n2h_dat.sh) [//]: # (```) ### 2. Testing and Inference **a. Video Processing:** #### 1. Configure the script file: Open and edit the file `test_scripts/ast_inference_video.sh`. Here, you need to provide the path to the trained checkpoint and the input/output video paths. #### 2. Run the video processing script: After completing the configuration, navigate to the project’s root directory and execute the following command: ```shell bash test_scripts/ast_inference_video.sh ``` **b. Image Directory Processing:** #### 1. Configure the script file: Open and edit the file `test_scripts/ast_n2h_dat.sh`. Here, you need to provide the path to the trained checkpoint and the input/output image directory paths. #### 2. Run the image directory processing script: After completing the configuration, navigate to the project’s root directory and execute the following command: ```shell bash test_scripts/ast_n2h.sh ``` ## Citation [//]: # (```bibtex) [//]: # (@article{crnet2025,) [//]: # ( title={CR-Net: A Continuous Rendering Network for Improving Robustness to Low-illumination},) [//]: # ( author={},) [//]: # ( journal={},) [//]: # ( year={2025}) [//]: # (}) [//]: # (```) ## References 1. https://github.com/EndlessSora/TSIT 2. https://github.com/astra-vision/CoMoGAN 3. https://github.com/AlienZhang1996/S2WAT ## License This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.