I am following this tutorial: https://blog.paperspace.com/mask-r-cnn-in-tensorflow-2-0/ in order to train a custom dataset for object detection. When I run the code for training (under paragraph: "Train Mask R-CNN in TensorFlow 1.0"), I get this error on colab:
NameError Traceback (most recent call last)
<ipython-input-31-794112aa6465> in <module>()
6 import mrcnn.config
7
----> 8 import mrcnn.model
9
10 class KangarooDataset(mrcnn.utils.Dataset):
/content/drive/MyDrive/How_to_Train_an_Object_Detection_Model_with_Keras/Mask_RCNN/mrcnn/model.py in <module>()
255
256
--> 257 class ProposalLayer(KE.Layer):
258 """Receives anchor scores and selects a subset to pass as proposals
259 to the second stage. Filtering is done based on anchor scores and
NameError: name 'KE' is not defined
After searching I tried to check that RCNN is ok with this: Import Matterport's Mask-RCNN model from github - error:ZipImportError: bad local file header with the solution that the guy in the end suggests. I have also found this: NameError: name 'K' is not defined so I tried this command:
from keras import backend as KE
(instead of K, I put KE) but it didn't work!
Do you have any idea how to fix that error?
Ok, I tried this github repository instead the original MaskRCNN: https://github.com/akTwelve/Mask_RCNN with the latest tensorflow (2.7.0) + Keras (2.7.0) installed on colab. It seems to overcome the above problem I described...I do not know why..!