search

marcellacornia / mlnet

A Deep Multi-Level Network for Saliency Prediction. ICPR 2016
MIT License
94 stars 37 forks source link

Errors using the pre-trained model #4

Closed guiluz closed 7 years ago

guiluz commented 7 years ago

Hi Marcella, I tried to compute saliency maps using the pre-trained model with the MIT300 benchmark but I have some errors. I changed the image resolution in the config.py, however I don't know if I need to add anything more. I don't know if there are problems for specific software versions, I have Anaconda Python 3.5.2 and Keras 1.2.0. And, I changed the backend to Theano already.

Here are the massages I got when I run python main.py test "image_folder"

Load weights ML-Net Predict saliency maps for C:\Users\Guilherme\Desktop\BenchmarkIMAGES Exception in thread Thread-37: Traceback (most recent call last): File "C:\Program Files\Anaconda3\lib\threading.py", line 914, in _bootstrap_inner self.run() File "C:\Program Files\Anaconda3\lib\threading.py", line 862, in run self._target(*self._args, **self._kwargs) File "C:\Program Files\Anaconda3\lib\site-packages\keras\engine\training.py", line 409, in data_generator_task generator_output = next(generator) File "main.py", line 36, in generator_test yield preprocess_images(images[counter:counter + b_s], shape_r, shape_c) File "C:\Users\Guilherme\Desktop\MLNET\utilities.py", line 36, in preprocess_images padded_image = padding(original_image, shape_r, shape_c, 3) File "C:\Users\Guilherme\Desktop\MLNET\utilities.py", line 11, in padding original_shape = img.shape AttributeError: 'NoneType' object has no attribute 'shape'

Traceback (most recent call last): File "main.py", line 71, in predictions = model.predict_generator(generator_test(b_s=1, imgs_test_path=imgs_test_path), nb_imgs_test) File "C:\Program Files\Anaconda3\lib\site-packages\keras\engine\training.py", line 1729, in predict_generator outs = self.predict_on_batch(x) File "C:\Program Files\Anaconda3\lib\site-packages\keras\engine\training.py", line 1319, in predict_on_batch self.internal_input_shapes) File "C:\Program Files\Anaconda3\lib\site-packages\keras\engine\training.py", line 74, in standardize_input_data 'Found: ' + str(data)[:200] + '...') TypeError: Error when checking : data should be a Numpy array, or list/dict of Numpy arrays. Found: None...

baraldilorenzo commented 7 years ago

Hi @guiluz, it seems that you are not loading the input images correctly (indeed, your img variable is None). Did you add the final slash in the second argument of main.py?

e.g. (from the readme) python main.py test path/to/images/folder/

guiluz commented 7 years ago

Hi @baraldilorenzo,

Of course! I was doing it wrong. Thanks for your help.