BKHMSI/Sketchback
1
1
Fork 0
mirror of https://github.com/BKHMSI/Sketchback.git synced 2024-11-20 13:02:37 +03:00
Code Issues Projects Releases Wiki Activity

ReadMe draft 4

Browse Source
This commit is contained in:
Badr AlKhamissi 2017-06-02 10:55:32 +02:00 committed by GitHub
parent 53c1fb9742
commit 987087bbbc
1 changed files with 20 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
README.md
View File

@ -1,5 +1,5 @@
# Sketchback: Convolutional Sketch Inversion using Keras
Implementation of sketch inversion using deep convolution neural networks (_synthesising photo-realistic images from pencil sketches_) following the work of [Convolutional Sketch Inversion][conv_paper] and [Scribbler][scribbler].
Keras implementation of sketch inversion using deep convolution neural networks (_synthesising photo-realistic images from pencil sketches_) following the work of [Convolutional Sketch Inversion][conv_paper] and [Scribbler][scribbler].
We focused on sketches of human faces and architectural drawings of buildings. However according to [Scribbler][scribbler] and our experimentation with their proposed framework, we believe that given a large dataset and ample training time, this network could generalize to other categories as well.
@ -7,13 +7,26 @@ We trained the model using a dataset generated from a large database of face ima
## Results
### Faces
<p align="center" width="500" height="165"><img src="Examples/faces_1.png"></p>
<p align="center" width="500" height="165"><img src="Examples/faces_2.png"></p>
<p align="center" width="500" height="165"><img src="Examples/faces_4.png"></p>
<table>
<tr>
<td><p align="center" width="500" height="165"><img src="Examples/faces_1.png"></p></td>
<td><p align="center" width="500" height="165"><img src="Examples/faces_2.png"></p></td>
</tr><tr>
<td><p align="center" width="500" height="165"><img src="Examples/faces_3.png"></p></td>
<td><p align="center" width="500" height="165"><img src="Examples/faces_4.png"></p></td>
</tr>
</table>
### Buildings
<p align="center"><img src="Examples/building_7.png"></p>
<p align="center"><img src="Examples/building_9.png"></p>
<table>
<tr>
<td><p align="center"><img src="Examples/building_7.png"></p></td>
<td><p align="center"><img src="Examples/building_8.png"></p></td>
</tr><tr>
<td><p align="center"><img src="Examples/building_9.png"></p></td>
<td><p align="center"><img src="Examples/coliseum.png"></p></td>
</tr>
</table>
## Datasets
We used the following datasets to train, validate and test our model:
@ -33,6 +46,7 @@ Furthermore, due to the low number of images of buildings available, we applied
## Network Architecture
<p align="center"><img src="Examples/scibbler_architecture.png"></p>
We used the network architecture used in [Scribbler][scribbler]. The generator follows an encoder-decoder design, with down-sampling steps, followed by residual layers, followed by up-sampling steps.
## Loss Functions