Author(s):
Tiangxiang Gao
Description:
First, we use feature pyramid encoding geometry to initially extract the features of the image. Next, we introduced attention blocks to measure similarity between images. Then, we propose a hybrid cascade structure of resolution-boosting stages and interval-reducing stages. This structure can significantly reduce computing time and memory consumption, while obtaining high reconstruction accuracy, especially for the edges of objects
Method parameters:
image resolution 1088*1920
Used image set:
Reconstructions from provided image set
Publication:
Software or method website:
Error Visualization:
| Precision | Recall |