Daniele Picone
commited on
Commit
·
7e7ca5d
1
Parent(s):
0665278
Updated readme
Browse files
README.md
CHANGED
|
@@ -1,3 +1,46 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-4.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-4.0
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
# Description
|
| 6 |
+
|
| 7 |
+
Series of weights recovered by training the `ligthning` torch model
|
| 8 |
+
UnrolledSystem implemented in:
|
| 9 |
+
(Unrolled demosaicking)[https://github.com/mattmull42/unrolled_demosaicking]
|
| 10 |
+
|
| 11 |
+
The network was trained over 15 color filter array patterns:
|
| 12 |
+
- `bayer`
|
| 13 |
+
- `binning`
|
| 14 |
+
- `chakrabarti`
|
| 15 |
+
- `gindele`
|
| 16 |
+
- `hamilton`
|
| 17 |
+
- `honda`
|
| 18 |
+
- `honda2`
|
| 19 |
+
- `kaizu`
|
| 20 |
+
- `kodak`
|
| 21 |
+
- `luo`
|
| 22 |
+
- `quad_bayer`
|
| 23 |
+
- `random`
|
| 24 |
+
- `sparse_3`
|
| 25 |
+
- `wang`
|
| 26 |
+
|
| 27 |
+
The network is based on U-NET, unrolled over 4 stages, and plugged into an ADMM solver. The network is trained over 300 natural images, cut into patches of size 64 x 64.
|
| 28 |
+
|
| 29 |
+
The three versions given in this repository are:
|
| 30 |
+
- `4`: Baseline weights.
|
| 31 |
+
- `4B`: Variant with different training set.
|
| 32 |
+
- `4V`: Introduces geometric transformations on the patterns.
|
| 33 |
+
|
| 34 |
+
# Citation
|
| 35 |
+
|
| 36 |
+
If you use this dataset, please cite:
|
| 37 |
+
```bibtex
|
| 38 |
+
@InProceedings{muller_eusipco_2024,
|
| 39 |
+
author = {Muller, Matthieu and Picone, Daniele and Dalla Mura, Mauro and Ulfarsson, Magnus Orn},
|
| 40 |
+
booktitle = {European Signal Processing Conference ({EUSIPCO})},
|
| 41 |
+
title = {Pattern-invariant unrolling for robust demosaicking},
|
| 42 |
+
year = {2024},
|
| 43 |
+
}
|
| 44 |
+
```
|
| 45 |
+
|
| 46 |
+
|