UNSUPERVISED IMAGE DENOISING USING DEEP LEARNING ( MONTE CARLO SURE )
Please find the Model.
Please find the Dataset .
Please find the Final Report
Certainly! Here's an updated version of your key feature section for your Unsupervised Image Denoising project, emphasizing the improvements related to loss and comparing it to the DNCNN neural network:
Key Features:
Implemented data augmentation techniques to increase the diversity of training samples, enhancing the robustness and generalization of our unsupervised image denoising model. This allowed our model to perform effectively across a wide range of noisy scenarios.
Conducted an extensive hyperparameter tuning process to optimize model settings, ensuring our unsupervised image denoising algorithm performs at its peak efficiency. Fine-tuning hyperparameters improved the model's convergence speed and overall performance.
Leveraged pre-trained models and fine-tuning to harness the power of prior knowledge, significantly improving the performance of our unsupervised image denoising project. This enabled our model to capture essential features from a diverse range of images, leading to better denoising results.
Introduced architectural changes TO U-NET that led to an improved loss function. By modifying the model's architecture, we achieved a better loss compared to the DNCNN neural network. The modified architecture allowed our model to capture noise characteristics more effectively, resulting in enhanced denoising performance.
Unsupervised Image Denoising using Monte Carlo Sure (SURE) Loss. Our goal is to enhance image quality by removing noise from images without the need for ground truth data. Leveraging the SURE loss, our approach ensures robust denoising across various noisy conditions, contributing to real-world image processing solutions.