Super Resolution

This is the Weights & Biases community benchmark for enhancing low resolution images with deep learning. The goal is to predict the most realistic high resolution image given an input image that's 8x less resolution.

Getting started

Follow the instructions in the starter GitHub repository to upload your experiment runs to W&B. You can change the existing hyperparameters and architecture to see if you can improve the model's performance. You could try using an RNN, changing the loss function, augmenting the training data, and so on.

Evaluation

We use a perceptual distance metric (val_perceptual_distance) on the validation set to rank results (lower values are better). The starter repository automatically generates this metric.

How to submit your results

You can submit runs to this benchmark from the "Runs" table in your "Project workspace" tab. To submit a specific run, hover over the run's name, click on the three-dot menu icon that appears to the left of the name, and select "Submit to benchmark". All submissions are reviewed by the benchmark administrators before acceptance.