Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Releases: PawelKinczyk/AECVision

v0.4.0

03 Aug 21:07
@PawelKinczyk PawelKinczyk
ed17d61
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Verified
Learn about vigilant mode.

Choose a tag to compare

What's Changed

Full Changelog: v0.3.0...v0.4.0

Contributors

PawelKinczyk
Assets 2
Loading

v0.3.0

26 Jul 20:28
@PawelKinczyk PawelKinczyk
febc5a3
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Verified
Learn about vigilant mode.

Choose a tag to compare

What's Changed

  • Add change RGB to grayscale cells by @PawelKinczyk in #7
  • Add grayscale models by @PawelKinczyk in #8
  • Two new models! Train on the same dataset but in grayscale. Below comparison on test data:

12 classes model

Main conclusion is that using gray images to train and detection increased the measures of small groups like "stairs", "refrigerator", "dishwasher" and "table".

Precision-Confidence Curve
a) RGB
image

b) Gray
image

Confusion matrix
a) RGB
image

b) Gray
image

F1 curve
a) RGB
image

b) Gray
image

Walls

Better results at 0.1 confidence. Interesting peak near 1.0 confidence in model_walls_gray.

Precision-Confidence Curve
a) RGB
image

b) Gray
image

Full Changelog: v0.2.0...v0.3.0

Contributors

PawelKinczyk
Loading

v0.2.0

11 Jul 20:07
@PawelKinczyk PawelKinczyk
c4d7159
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Verified
Learn about vigilant mode.

Choose a tag to compare

What's Changed

New Contributors

Full Changelog: v0.1.0...v0.2.0

New wall detection model

I decided to train new model with less classes. I didn't change dataset since last release so i focus on biggest class. As i mention in readme about first model: "Walls are overrepresented because this is normal quantities in architectural plans."
image
This should improve of walls detection and YES! i get better results for walls with shorter train time (but train process could be longer because model is getting better even after 194 epoch). This is plot with "Train loss" :
image
And "Validation loss":
image

Validation

Confusion matrix:
image
Precision-Confidence curve:
image

Contributors

dependabot and PawelKinczyk
Loading

v0.1.0

11 Jul 18:03
@PawelKinczyk PawelKinczyk
7981a53
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Verified
Learn about vigilant mode.

Choose a tag to compare

This is the first version of AECVision project. Available one model which detects 12 classes:

  • 'bath',
  • 'dishwasher / washing machine',
  • 'door',
  • 'furniture',
  • 'refrigerator',
  • 'sink',
  • 'stairs',
  • 'stove',
  • 'table',
  • 'wall',
  • 'wc',
  • 'window'
Loading

AltStyle によって変換されたページ (->オリジナル) /