kurt willis
commited on
exchange images
Browse files
README.md
CHANGED
|
@@ -7,9 +7,8 @@
|
|
| 7 |
In order to address camera hardware-drift we require two ingredients: raw sensor data and an image processing model. This code repository contains the materials for the second ingredient, the image processing model, as well as scripts to load lada and run experiments. For a conceptual overview of the project we reocommend the [project site](https://aiaudit.org/lens2logit/) or the [full paper](https://openreview.net/forum?id=DRAywM1BhU).
|
| 8 |
|
| 9 |
## A short introduction
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
</p>
|
| 13 |
|
| 14 |
To create an image, raw sensor data traverses complex image signal processing pipelines. These pipelines are used by cameras and scientific instruments to produce the images fed into machine learning systems. The processing pipelines vary by device, influencing the resulting image statistics and ultimately contributing to what is known as hardware-drift. However, this processing is rarely considered in machine learning modelling, because available benchmark data sets are generally not in raw format. Here we show that pairing qualified raw sensor data with an explicit, differentiable model of the image processing pipeline allows to tackle camera hardware-drift.
|
| 15 |
|
|
@@ -121,9 +120,8 @@ foo@bar:~$ python train.py \
|
|
| 121 |
```
|
| 122 |
## Virtual lab log
|
| 123 |
We maintain a collaborative virtual lab log at [this address](http://deplo-mlflo-1ssxo94f973sj-890390d809901dbf.elb.eu-central-1.amazonaws.com/#/). There you can browse experiment runs, analyze results through SQL queries and download trained processing and task models.
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
</p>
|
| 127 |
|
| 128 |
### Review our experiments
|
| 129 |
Experiments are listed in the left column. You can select individual runs or compare metrics and parameters across different runs. For runs where we tracked images of intermediate processing steps and images of the gradients at these processing steps you can find at the bottom of a run page in the *results* folder for each epoch.
|
|
|
|
| 7 |
In order to address camera hardware-drift we require two ingredients: raw sensor data and an image processing model. This code repository contains the materials for the second ingredient, the image processing model, as well as scripts to load lada and run experiments. For a conceptual overview of the project we reocommend the [project site](https://aiaudit.org/lens2logit/) or the [full paper](https://openreview.net/forum?id=DRAywM1BhU).
|
| 8 |
|
| 9 |
## A short introduction
|
| 10 |
+
|
| 11 |
+
|
|
|
|
| 12 |
|
| 13 |
To create an image, raw sensor data traverses complex image signal processing pipelines. These pipelines are used by cameras and scientific instruments to produce the images fed into machine learning systems. The processing pipelines vary by device, influencing the resulting image statistics and ultimately contributing to what is known as hardware-drift. However, this processing is rarely considered in machine learning modelling, because available benchmark data sets are generally not in raw format. Here we show that pairing qualified raw sensor data with an explicit, differentiable model of the image processing pipeline allows to tackle camera hardware-drift.
|
| 14 |
|
|
|
|
| 120 |
```
|
| 121 |
## Virtual lab log
|
| 122 |
We maintain a collaborative virtual lab log at [this address](http://deplo-mlflo-1ssxo94f973sj-890390d809901dbf.elb.eu-central-1.amazonaws.com/#/). There you can browse experiment runs, analyze results through SQL queries and download trained processing and task models.
|
| 123 |
+
|
| 124 |
+
|
|
|
|
| 125 |
|
| 126 |
### Review our experiments
|
| 127 |
Experiments are listed in the left column. You can select individual runs or compare metrics and parameters across different runs. For runs where we tracked images of intermediate processing steps and images of the gradients at these processing steps you can find at the bottom of a run page in the *results* folder for each epoch.
|