Problem: Learn and reconstruct a set of images as any resolution.
Method: Using multi-resolutional dense grid based bilinear resampling and neural networks.
- Contrusct a multi-resolutional dense grid as coordinate-based data structures to store learnable features.
- Each pixel of image is transformed into that coordinate system (e.g. UV).
- Color value is determined by bilinear interpolation of features stored in neighbor vertices.
- Training and testing samples are generated from the pixels of original images.
- Samples for reconstruction can be generated at any resolution.
Details: notebook/demo.ipynb
- Anaconda, for python package management.
conda env create -f environment.yaml
conda activate dgir