Thanks to gradient checkpointing, the framework is trainable end-to-end with limited memory resources and weak (image-level) supervision, allowing for the joint optimization of all parameters.

This article focuses on end-to-end image matching through joint key-point detection and descriptor extraction. To find repeatable and high discrimination key points, we improve the deep matching ...

Download the test images, OR export a WebP image from Krita that has a transparent background. In the export settings, enable 'Lossless encoding' and set the quality to 100%.