-
Notifications
You must be signed in to change notification settings - Fork 16
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add python version of tilers #93
Conversation
@Wovchena please have a look at python stuff. IS tiler is not fully finished (XAI is lacking). Also, tests are missing. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
All in all the design looks good
map_pixel = cls_map[hi, wi] | ||
merged_pixel = merged_map[class_idx][y_1 + hi, x_1 + wi] | ||
if merged_pixel != 0: | ||
merged_map[class_idx][y_1 + hi, x_1 + wi] = 0.5 * (map_pixel + merged_pixel) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It seems that the first elements of saliency_maps[1:]
are going to have smaller weights compared to the last elements if more that 2 maps overlap at any pixel. The last elem has weight 0.5, the previous 0.25 etc. Is this intentional?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, that's a heuristic for merging. Max looks more natural, but during the initial experiments it produced more artifacts. I'll have a look at this one more time.
@Wovchena please, have a look at tests implementation. CPP tests currently just skip tiling models. |
What does this PR do?
Fixes # (issue)
Before submitting