You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@eldar & @andreas-eberle
Hi, I try to freeze your model by the following code:
restorer = tf.train.Saver()
sess = tf.Session()
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
##restore variables from disk ckpt file.
restorer.restore(sess, "model/mpii/mpii-single-resnet-101")
#write graph
tf.train.write_graph(sess.graph_def, './', 'graph_unfreeze.pb', as_text=True)
The "frozen_graph.pb" file is generated successfully. But when I use the "frozen_graph.pb" file to estimatie body part, the result is wrong, and I don't know why.
Can you help me to export the graph?
Thank you very much~~
The text was updated successfully, but these errors were encountered:
@zjbaby How did you know the output nodes ('pose/part_pred/block4/BiasAdd,pose/locref_pred/block4/BiasAdd) ??
I am trying to convert the model to onnx but unable to convert as a conversion tool is not getting the output node recognized. Please assist.
@eldar & @andreas-eberle
Hi, I try to freeze your model by the following code:
restorer = tf.train.Saver()
sess = tf.Session()
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
##restore variables from disk ckpt file.
restorer.restore(sess, "model/mpii/mpii-single-resnet-101")
#write graph
tf.train.write_graph(sess.graph_def, './', 'graph_unfreeze.pb', as_text=True)
#freeze graph
freeze_graph.freeze_graph('./graph_unfreeze.pb', '', False, "model/mpii/mpii-single-resnet-101", 'pose/part_pred/block4/BiasAdd,pose/locref_pred/block4/BiasAdd', None, None, './frozen_graph.pb', False, "")
The "frozen_graph.pb" file is generated successfully. But when I use the "frozen_graph.pb" file to estimatie body part, the result is wrong, and I don't know why.
Can you help me to export the graph?
Thank you very much~~
The text was updated successfully, but these errors were encountered: