tabayashi0117 / score-cam Goto Github PK

Hello. I like the clean implementation of yours. Currently there is no license for the repository. Would you kindly add one, for the community to modify and reuse the code?

can you share the requirements.txt file? I am having some errors.

one error this cell

this error

Originally posted by @en-ashay in #1 (comment)

Hello.
First of all, I appreciate you for providing such a wonderful code.
I have a question in your code.
More specifically, about function ScoreCam() in gradcamutils.py.

Could you explain why you choose activation maps with large standard deviations?
I mean the procedure of extracting effective maps in case of max_N != =1.

How can Faster ScoreCAM be used in a medical images classification app, so that the classified images appear with heatmaps?

cv.resize does not work as expected in this line. Please, rewrite this line as:

act_map_resized_list = [np.transpose(cv2.resize(act_map_array[0,:,:,k], input_shape[:2], interpolation=cv2.INTER_LINEAR)) for k in range(act_map_array.shape[3])]

otherwise error appears for rectangle image shape in line: below

Hello
I am training VGG16 on my own dataset with TF2.2. The Score-CAM part seemed fine. But when i working on the Guided Score-CAM part, it raise the error 'Tensor is not an element of this graph.'
Here's the code:

`

layer_name = 'block5_conv3'

img_path = '1.jpg'

img_array = read_and_preprocess_img(img_path, size=(224,224))

model = load_model('my_model.h5')

guided_model = build_guided_model(model)

tf.compat.v1.disable_eager_execution()

saliency = GuidedBackPropagation(guided_model, img_array, layer_name)

`

Error screenshot:

This is my model:
`conv_base = VGG16(include_top=False, weights='imagenet',input_shape=(224, 224, 3))

top_layer = Flatten()(conv_base.output)

top_layer = Dense(256, activation='relu')(top_layer)

top_layer = Dense(3, activation='softmax')(top_layer)

model = Model(inputs=conv_base.input, outputs=top_layer)

model.compile(optimizer=SGD(lr=1e-4), loss='categorical_crossentropy', metrics=['accuracy'])`

Any help would be appreciated, thanks!

ValueError Traceback (most recent call last)
in ()
7 img_array = read_and_preprocess_img('/content/drive/MyDrive/Phd_Data_set_Implementation/Aug-Training-four/BL/HL_0_1013.jpeg', size=(224,224))
8
----> 9 score_cam = ScoreCam(model,img_array,layer_name)
10
11 plt.imshow(score_cam)

2 frames
in ScoreCam(model, img_array, layer_name, max_N)
5 def ScoreCam(model, img_array, layer_name, max_N=-1):
6
----> 7 cls = np.argmax(model.predict(img_array))
8 act_map_array = Model(inputs=model.input, outputs=model.get_layer(layer_name).output).predict(img_array)
9

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py in predict(self, x, batch_size, verbose, steps, callbacks, max_queue_size, workers, use_multiprocessing)
1575 """
1576 base_layer.keras_api_gauge.get_cell('predict').set(True)
-> 1577 version_utils.disallow_legacy_graph('Model', 'predict')
1578 self._check_call_args('predict')
1579 _disallow_inside_tf_function('predict')

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/version_utils.py in disallow_legacy_graph(cls_name, method_name)
120 " call {cls_name}.{method_name} with eager mode enabled.")
121 error_msg = error_msg.format(cls_name=cls_name, method_name=method_name)
--> 122 raise ValueError(error_msg)
123
124

ValueError: Calling Model.predict in graph mode is not supported when the Model instance was constructed with eager mode enabled. Please construct your Model instance in graph mode or call Model.predict with eager mode enabled.