VNect-tensorflow issue using converted vnect.pkl

i converted the caffemodel using the caffe_weights_to_pickle.py. however, when i want to load the vnect.pkl with tensorflow, i get: DataLossError (see above for traceback): Unable to open table file models/weights/vnect_tf: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?

thanks for any help

May 29 '18 08:05 mlasy

check vnect_model.py in the models folder. There's a function called "load_weights" and you could use saver.save to get vnect_tf.

Jun 06 '18 03:06 yukaliu07

@yukaliu07 could you maybe provide some more detail on how to use saver.save? I have implemented it as follows, but the model files that this returns, give me wrong results:

with tf.Session() as sess:
    saver = tf.train.Saver()
    tf_writer = tf.summary.FileWriter(logdir='./', graph=sess.graph)

    sess.run(tf.global_variables_initializer())
    saver.save(sess, './vnect_tf')
    print(model.res5b_branch2c_new)
    print(model.heatmap, model.x_heatmap, model.y_heatmap, model.z_heatmap)

Jun 27 '18 07:06 lavage

@lavage with tf.Session() as sess: saver = tf.train.Saver() model.load_weights(sess, model_file) save_path = saver.save(sess, "./vnect_tf")

Jul 02 '18 02:07 yukaliu07

@yukaliu07 @lavage Thank you so much!! It perfectly solved my problem. :D

Jul 03 '18 05:07 powermew

@yukaliu07 I use the code but the pkl can't convert correctly. Error message: Traceback (most recent call last): File "models/vnect_model.py", line 175, in model = VNect(368) File "models/vnect_model.py", line 13, in init self._create_network() File "models/vnect_model.py", line 94, in _create_network self.res4d_branch2a = tc.lvnectayers.conv2d(self.res4c, kernel_size=1, num_outputs=256, scope='res4d_branch2a') AttributeError: 'module' object has no attribute 'lvnectayers'

Aug 25 '18 01:08 fenglupeter

@yukaliu07 @powermew @lavage

Hello, I had tried this code with these follows.

1.Download weights from the author of paper
2.change weights with using './caffe_weight_to_pickle.py'

run '/models/vnect_model.py' to get

vnect_tf.data-00000-of-00001 vnect_tf.index vnect_tf.meta

run '/demo_tf.py'

But I am getting an error like this 2018-09-27 16:11:48.784541: W tensorflow/core/util/tensor_slice_reader.cc:95] Could not open models/weights/vnect_tf: Failed precondition: models/weights/vnect_tf; Is a directory: perhaps your file is in a different file format and you need to use a different restore operator?

I have tried editing vnect_model.py file as per

@lavage with tf.Session() as sess: saver = tf.train.Saver() model.load_weights(sess, model_file) save_path = saver.save(sess, "./vnect_tf")

But I couldn't solve the problem, Can someone help me with this

Sep 27 '18 14:09 Robinchacko

@Robinchacko At step 3 on your method, i guess you should fix the location of model file in demo_tf.py

Sep 28 '18 01:09 powermew

I changed code line 173-181 in vnect_tf.py

if __name__ == '__main__':
    model_file = 'vnect.pkl'
    model = VNect(368)

    with tf.Session() as sess:
        saver = tf.train.Saver()
        model.load_weights(sess, model_file)
        save_path = saver.save(sess, "./vnect_tf")

And also changed line 21 in demo_tf.py

parser.add_argument('--model_file', default='models/weights/vnect_tf')

Sep 28 '18 05:09 powermew

can any one provide model files please? [email protected] , author is not replying.