Open oplatek opened 4 years ago
the disadvantages you mention are quite heavy and probably making unfeasible to obtain an solution to the problem. This seems more an unsupervised problem and probably stuff like creating self attention maps could help.
Actually found this interesting post using it already for xray images https://towardsdatascience.com/self-attention-in-computer-vision-2782727021f6
@anguelos @ducha-aiki any experience with self-attention ?
For explanations I'd go for this: https://arxiv.org/abs/2001.08593
For training: https://github.com/sdoria/SimpleSelfAttention
This proposal suggests a principal solution to build a model that will predict Bounding Box (BB) around lungs, artifacts and other well-defined objects on the images.
The COVID19 and other x-ray chest datasets contain chest X-ray images capturing not only Lungs but sometimes also arms, neck, etc. The images also contain text labels and other artifacts.
The advantages:
The disadvantages:
Example images with textual artifacts, "pipes in lungs artifact", the edge of image artifact, ..
![lungs-artifacts2](https://user-images.githubusercontent.com/229266/77844992-62d9ab00-71ab-11ea-962f-10eaa43ff46c.png)