I am experiencing an issue when i test object detection or when i train
this error keeps occuring:
Traceback (most recent call last):
File "object_detection/builders/model_builder_test.py", line 23, in
from object_detection.builders import model_builder
File "/root/dia/tensorflow/models/research/object_detection/builders/model_builder.py", line 56, in
from object_detection.models.ssd_mobilenet_v3_feature_extractor import SSDMobileNetV3LargeFeatureExtractor
File "/root/dia/tensorflow/models/research/object_detection/models/ssd_mobilenet_v3_feature_extractor.py", line 25, in
from nets.mobilenet import mobilenet_v3
ImportError: cannot import name 'mobilenet_v3'
i have no problem with path or anything
cuda version : CUDA Version 9.0.176
tensorflow version 1.12 and 1.14 (tried with both)
python version (3.5,3.6,3.7) tried them all
Same here. Using TF.1.13.1
Have it fixed. Check your PYTHONPATH
Have it fixed. Check your PYTHONPATH
i checked and everything alright
its not the first installation of object detection api i make
this is weird this time
i guess it's due to the last commit of 15-october
i fixed that error by git checkout "ID"
ID of the commit before adding mobilenet_v3
now it works perfectly
So may I ask that how did you fix that?
i fixed that error by git checkout "ID"
ID of the commit before adding mobilenet_v3
now it works perfectly
can you explain more, what do you mean with gut checkout 'ID' ?
For future readers, I've faced the same problem. Turns out it's because the nets folder is probably outdated In the version you're using.
A quick workaround is to download the latest version of TensorFlow models.
https://github.com/tensorflow/models
then copy contents of "/models/research/slim/nets/mobilenet" from the latest version to your project directory following a similar path, must be "/models/research/slim/nets/mobilenet" again. Just replace the files when prompted, and you've gotta build slim after this copy.
in the folder dir "~/model/research/slim"
python setup.py build
python setup.py install
Worked like a charm for me.
PS: Also make sure your PYTHONPATHs are all good.