# Denoising without Clean and pair data [Self-Supervised Denoising](https://arxiv.org/pdf/1901.10277.pdf)  **Framework**  **Practical Framework** - Why Self-Supervision ? - To denoise an image without clean and pair data - How to train the network ? - Using rotated images as input - Making half-plane receptive field - Restricting receptive field by using feature-level offset in layer by layer <img width="939" alt="image" src="https://user-images.githubusercontent.com/59157285/113590801-0e92c200-966e-11eb-92d0-d2346131408e.png"> - Training & Inference - **Training**   - **Inference**  - **Summary**    