Error when training CoarseNet_train.py

[loverpingguo]

when I training CoarseNet_train.py, （the image size is 512*512 ，batchsize=2 ），The following error occurred：tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor with shape[2,64,512,512] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
GPU ：NVIDIA GeForce GTX TITAN X

[luannd]

This error is the result of not having enough GPU memory. What is your GPU memory size? I suggest you use small image size to train or set batchsize to 1 if you prefer big size input.

[loverpingguo]

12GB

[loverpingguo]

Does training make sense when set batchsize to 1?

[luannd]

It depends on your goal. You can reduce input size (and increase batch size) by cropping your images.

[loverpingguo]

I can't open the link for the pretrained models for CoarseNet and FineNet ?

[loverpingguo]

what the image size is , you used when you were training

[luannd]

As specified in the code, input size is 400 but you can use any size which is suitable for your machine.
I already created model links in Dropbox besides Google drive:
CoarseNet: https://www.dropbox.com/s/gppil4wybdjcihy/CoarseNet.h5?dl=0
FineNet: https://www.dropbox.com/s/k7q2vs9255jf2dh/FineNet.h5?dl=0

[loverpingguo]

I have cropped image to 400*400,but when training,and running tolabel2mnt in evaluate_trainingThe following error occurred：
File "d:\wu\MinutiaeNet\CoarseNet\CoarseNet_utils.py", line 289, in label2mnt
mnt_list = np.array(zip(mnt_sparse.row, mnt_sparse.col), dtype=np.int32)
TypeError: int() argument must be a string, a bytes-like object or a number, not 'zip'

[wajihullahbaig]

That would be a Python 3 error as the code is for python 2. zip function returns generators.
You will need to change to code to the following:
mnt_list = np.array([*map(tuple,zip(mnt_sparse.row, mnt_sparse.col))])

You can also check my fork for Python3 version of the code.

[mimrance]

I am running the demo_CoarseNet notebook on pyhton 3 by incorporating changes as the uploaded code was for python 2. i am getting following error:

---

IndexError Traceback (most recent call last)
Cell In[3], line 15
10 mkdir(output_dir + '/' + set_name + '/OF_results/')
12 logging.info("Predicting "%s":" % (set_name))
---> 15 main_net_model = CoarseNetmodel((None, None, 1), CoarseNet_path, mode='deploy')
17 # ====== Load FineNet to verify
18 if isHavingFineNet == True:

File ~/FingerprintImageQuality-main/MiDeCon/CoarseNet/CoarseNet_model.py:168, in CoarseNetmodel(input_shape, weights_path, mode)
161 filter_img_real = Conv2D(filters_cos.shape[3],(filters_cos.shape[0],filters_cos.shape[1]),
162 weights=[filters_cos, np.zeros([filters_cos.shape[3]])], padding='same',
163 name='enh_img_real_1')(img_input)
164 filter_img_imag = Conv2D(filters_sin.shape[3],(filters_sin.shape[0],filters_sin.shape[1]),
165 weights=[filters_sin, np.zeros([filters_sin.shape[3]])], padding='same',
166 name='enh_img_imag_1')(img_input)
--> 168 ori_peak = Lambda(ori_highest_peak)(ori_out_1)
169 ori_peak = Lambda(select_max)(ori_peak) # select max ori and set it to 1
171 # Use this function to upsample image

File ~/anaconda3/envs/tenserflow_env/lib/python3.11/site-packages/keras/utils/traceback_utils.py:70, in filter_traceback..error_handler(*args, **kwargs)
67 filtered_tb = _process_traceback_frames(e.traceback)
68 # To get the full stack trace, call:
69 # tf.debugging.disable_traceback_filtering()
---> 70 raise e.with_traceback(filtered_tb) from None
71 finally:
72 del filtered_tb

File ~/FingerprintImageQuality-main/MiDeCon/CoarseNet/CoarseNet_utils.py:224, in ori_highest_peak(y_pred, length)
223 def ori_highest_peak(y_pred, length=180):
--> 224 glabel = gausslabel(length=length,stride=2).astype(np.float32)
225 y_pred = tf.convert_to_tensor(y_pred, np.float32)
226 ori_gau = K.conv2d(y_pred,glabel,padding='same')

File ~/FingerprintImageQuality-main/MiDeCon/CoarseNet/MinutiaeNet_utils.py:132, in gausslabel(length, stride)
130 delta = np.array(np.abs(label - y), dtype=int)
131 delta = np.minimum(delta, length-delta)+length/2
--> 132 return gaussian_pdf[delta]

IndexError: Exception encountered when calling layer "lambda_3" (type Lambda).

arrays used as indices must be of integer (or boolean) type

Call arguments received by layer "lambda_3" (type Lambda):
• inputs=tf.Tensor(shape=(None, None, None, 90), dtype=float32)
• mask=None
• training=None