Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Fix for TIKA-2400 Standardizing current Object Recognition REST parsers #208
This PR consists of,
How to test?
@ThejanW Thanks again for cleaning the code that evolved over past two years!
I made some suggestions.
Also, I cant test video docker
docker run -it -p 8764:8764 uscdatascience/inception-rest-tika Can't import video libraries, No video functionality is available Traceback (most recent call last): File "/usr/bin/inceptionapi.py", line 265, in <module> app = Classifier(__name__) File "/usr/bin/inceptionapi.py", line 221, in __init__ self.names = imagenet.create_readable_names_for_imagenet_labels() File "/models-c15fada28113eca32dc98d6e3bec4755d0d5b4c2/slim/datasets/imagenet.py", line 93, in create_readable_names_for_imagenet_labels assert num_synsets_in_ilsvrc == 1000 AssertionError
docker rmi -f for previous images but stil something is wrong.
I cant test image as well as video docker.
I will wait for others to test and confirm if the issue is with my docker setup or with the images
I was getting the same error. Nothing is wrong with your docker setup. The problem was with the download url of imagenet_lsvrc_2015_synsets.txt & imagenet_metadata.txt. Apparently tf maintainers have moved these meta files and models to another repo https://github.com/tensorflow/serving.
@thammegowda @chrismattmann @smadha This is complete now. I have updated tensorflow version and models to the latest(tf 1.4.0). Currently object rec REST parsers are not functioning due to the URL change of imagenet_lsvrc_2015_synsets.txt & imagenet_metadata.txt. By this PR, those issues can also be resolved. Therefore it would be nice if we can merge this before 1.17. Testing instructions are included in the initial comment.