A deep convolutional neural network which up-scales images without the typical pixelation artifacts.
Accepts JPEG and PNG files as inputs and outputs. However, PNG is recommended for both, as pre-existing JPEG artifacts may affect the quality of the up-scaling, and outputting as JPEG introduces unnecessary artifacts.
model: Select a network optimized for a specific input type: imagenet (natural images, default), anime (line art or stylized inputs), bilinear (fast low quality technique).
Many thanks to A. Nowell for the port to Algorithmia.