ValueError: Tried to convert 'shape' to a tensor and failed. Error: None values not supported.
To solve this for the mask RCNN
The solution I found and worked for me was from "mayurmahurkar commented on Dec 8, 2020"
PS: some of the functions lines are with slightly different names but they should be there in the model.py file
The following steps worked for me:
1. Upgrade the scripts by using the following line on the root folder:
tf_upgrade_v2 --intree Mask_RCNN --inplace --reportfile report.txt
This will automatically update the existing code to TF2. You will also get a list of changes made in report.txt
2. Replace the following line:
mrcnn_bbox = layers.Reshape((-1, num_classes, 4), name="mrcnn_bbox")(x)
with this this if-else code block:
if s[1]==None:
mrcnn_bbox = layers.Reshape((-1, num_classes, 4), name="mrcnn_bbox")(x)
else:
mrcnn_bbox = layers.Reshape((s[1], num_classes, 4), name="mrcnn_bbox")(x)
3. Change the following line:
indices = tf.stack([tf.range(probs.shape[0]), class_ids], axis=1)
with this line:
indices = tf.stack([tf.range(tf.shape(probs)[0]), class_ids], axis = 1)
4. Now, you need to replace:
from keras import saving
with:
from tensorflow.python.keras import saving
then you will also want to replace the lines in both if and else block:
saving.load_weights_from_hdf5_group(f, layers)
and so on with the follwoing lines, inside if and else block respectively:
saving.hdf5_format.load_weights_from_hdf5_group_by_name(f, layers)
saving.hdf5_format.load_weights_from_hdf5_group(f, layers)
Thanks to: @deluongo, @Trotts, @nielsuit227, @ibrahimLearning