How to make the model file, like CNNIQA-LIVE ? about cnniqa HOT 8 CLOSED

hello do you know the role of exp_id in the code ?

from cnniqa.

@RainFZY Read the Readme.md and use --help for these two *.py files to get help.

Just set --model_file, for example, python test_demo.py --im_path=data/I03_01_1.bmp --model_file=your/model/path.
It is the same in test_cross_dataset.py.

from cnniqa.

@salmora See #21 .

from cnniqa.

@RainFZY Read the Readme.md and use --help for these two *.py files to get help.

Just set --model_file, for example, python test_demo.py --im_path=data/I03_01_1.bmp --model_file=your/model/path.
It is the same in test_cross_dataset.py.

I mean how to create the model file itself. For example if i use CNN, then i have to create a model file CNNIQA-LIVE, which you already offered. But if I want switch to my own model, how can I get those model files? In other words, I know how to use them but i just have no idea how can I get them? Thanks a lot

from cnniqa.

@RainFZY Read the Readme.md and use --help for these two *.py files to get help.

Just set --model_file, for example, python test_demo.py --im_path=data/I03_01_1.bmp --model_file=your/model/path.
It is the same in test_cross_dataset.py.

model --> model file
CNN --> CNN-LIVE
other model --> ???
The ??? is what I need, i hope i can express my questions clearly this time.

from cnniqa.

@RainFZY I didn't get what you mean.
I suggest you read the whole repo again.

Based on your further comments, here I give more comments.
You said you have already run the training file main.py, so you would have get your trained model (line 174 in main.py saves the model with best val performance), which is named used the string trained_model_file = 'checkpoints/{}-{}-EXP{}-lr={}'.format(args.model, args.database, args.exp_id, args.lr), you can just rename the final model with a simple name, such as CNNIQA-[your_dataset_name]. Then you can use this final model in the test phase (Make sure you will not test the training images, since it does not make sense.).

from cnniqa.

I think in test_cross_dataset.py , they use another database just to see the performance of the modele , and in the paper they use TID2008 , you have to read the paper ,it describes this part.

from cnniqa.

@RainFZY I didn't get what you mean.
I suggest you read the whole repo again.

Based on your further comments, here I give more comments.
You said you have already run the training file main.py, so you would have get your trained model (line 174 in main.py saves the model with best val performance), which is named used the string trained_model_file = 'checkpoints/{}-{}-EXP{}-lr={}'.format(args.model, args.database, args.exp_id, args.lr), you can just rename the final model with a simple name, such as CNNIQA-[your_dataset_name]. Then you can use this final model in the test phase (Make sure you will not test the training images, since it does not make sense.).

Thanks a lot, my problem is solved.

from cnniqa.