1. Prerequisites:
python + python libraries [tensorflow (1.0.1 and above), numpy, scipy]
2. To run pre-trained models, use the following command:
python run_model.py model=<model> resolution=<resolution> use_gpu=<use_gpu>
where <model> is the corresponding phone model: {iphone, sony, blackberry}
<resolution>: {orig, high, medium, small, tiny}
<use_gpu>: {true, false}
example:
python run_model.py model=iphone resolution=orig use_gpu=true
3. Can I run the models on CPU?
Yes, you can set the parameter <use_gpu> to false, but note that in this case the performance will be low (up to 3-5 minutes per image)
4. What if I get an error: "OOM when allocating tensor with shape[...]"?
Your GPU does not have enough memory to process full-resolution images (this usually happens when it has less than 6GB or RAM)
Solutions:
a) Run the model on CPU (set the parameter <use_gpu> to false). Note that this will take considerably more time (3-5 minutes per image)
b) Use cropped images, set the parameter <resolution> to:
high - center crop of size 1680x1260 pixels
medium - center crop of size 1366x1024 pixels
small - center crop of size 1024x768 pixels
tiny - center crop of size 800x600 pixels
The less resolution is - the smaller part of the image will be processed
5. Folder structure:
models/ - pre-trained models for three phones (iphone, sony and blackberry)
test_photos/blackberry/ - test photos for blackberry phone
test_photos/iphone/ - test photos for iphone
test_photos/sony/ - test photos for sony phone
results/ - processed images will appear in this folder
run_model.py - python script that loads test images and applies pre-trained models to them
model.py - python script that specifies the architecture of the image enhancement network (resnet)
utils.py - python script for command arguments / input images processing
6. Any other questions / problems?
You can contact andrey.ignatoff@gmail.com for more information