problem with inference of covid-net-ct-a about covidnet-ct HOT 9 CLOSED

Hi Joachim,

The run_covidnet_ct.py script handles training, validation/testing and inference. As such, there is a required mode argument that should precede all other arguments, i,e,:
python run_covidnet_ct.py infer --input_height 512 --input_width 512 --auto_crop ...

One other thing to note is that the current COVIDNet-CT models take input images of size 512x512. You will get a shape-related error if you set --input_height 100 --input_width 100. These arguments are included in case we add new models with different input shapes, but for now they should be left as the default 512x512.

from covidnet-ct.

Hi Haydengunraj,
thanks for the quick help. I installed the new version and followed your advises, but get now a lot more errors. As I'm not very familiar with python, I need some help for just do inference with your models:

OUTPUT:
2020-09-17 08:41:19.634355: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
2020-09-17 08:41:19.639191: I tensorflow/core/common_runtime/process_util.cc:71] Creating new thread pool with default inter op setting: 4. Tune using inter_op_parallelism_threads for best performance.
Loading meta graph from C:/covid-net/COVIDNet-CT-master/COVIDNet-CT-A\model.meta
Traceback (most recent call last):
File "c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py", line 372, in
runner.infer(args.image_file, args.auto_crop)
File "c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py", line 244, in infer
graph, sess, saver = self.load_graph()
File "c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py", line 121, in load_graph
saver = tf.train.import_meta_graph(self.meta_file)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\training\saver.py", line 1435, in import_meta_graph
meta_graph_or_file, clear_devices, import_scope, **kwargs)[0]
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\training\saver.py", line 1457, in _import_meta_graph_with_return_elements
**kwargs))
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\meta_graph.py", line 806, in import_scoped_meta_graph_with_return_elements
return_elements=return_elements)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\util\deprecation.py", line 507, in new_func
return func(*args, **kwargs)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\importer.py", line 399, in import_graph_def
_RemoveDefaultAttrs(op_dict, producer_op_list, graph_def)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\importer.py", line 159, in _RemoveDefaultAttrs
op_def = op_dict[node.op]
KeyError: 'FusedBatchNormV3'

best regards
Joachim

from covidnet-ct.

Dear Haydengunraj,

thanks for your help. But unfortunately I run into other problems. I've downloaded the newest version. Must remark line 13 because it failed to load ConfusionMatrixDisplay.

python c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py infer --input_height 512 --input_width 512 --auto_crop --model_dir C:/covid-net/COVIDNet-CT-master/COVIDNet-CT-A --meta_name model.meta --ckpt_name model --image_file c:/test/a5ae5a83-a71d-4515-87df-6f15f691e166.jpg

gives now

2020-09-17 13:17:03.331029: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
2020-09-17 13:17:03.332444: I tensorflow/core/common_runtime/process_util.cc:71] Creating new thread pool with default inter op setting: 4. Tune using inter_op_parallelism_threads for best performance.
Loading meta graph from C:/covid-net/COVIDNet-CT-master/COVIDNet-CT-A\model.meta
Traceback (most recent call last):
File "c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py", line 373, in
runner.infer(args.image_file, args.auto_crop)
File "c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py", line 245, in infer
graph, sess, saver = self.load_graph()
File "c:/covid-net/COVIDNet-CT-master/run_covidnet_ct.py", line 122, in load_graph
saver = tf.train.import_meta_graph(self.meta_file)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\training\saver.py", line 1435, in import_meta_graph
meta_graph_or_file, clear_devices, import_scope, **kwargs)[0]
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\training\saver.py", line 1457, in _import_meta_graph_with_return_elements
**kwargs))
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\meta_graph.py", line 806, in import_scoped_meta_graph_with_return_elements
return_elements=return_elements)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\util\deprecation.py", line 507, in new_func
return func(*args, **kwargs)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\importer.py", line 399, in import_graph_def
_RemoveDefaultAttrs(op_dict, producer_op_list, graph_def)
File "C:\ProgramData\Anaconda3\lib\site-packages\tensorflow\python\framework\importer.py", line 159, in _RemoveDefaultAttrs
op_def = op_dict[node.op]
KeyError: 'FusedBatchNormV3'

from covidnet-ct.

Could you specify which TensorFlow version you're using? This may be due to a version mismatch.

from covidnet-ct.

I'm using Tensorflow 1.13.1 which is perfectly running with covid-net. Therefor I hesitate to upgrade

from covidnet-ct.

COVIDNet-CT has only been tested with TensorFlow 1.15, so it's difficult for me to say what the problem is. My best guess is that the FusedBatchNormV3 op in 1.15 is not supported by 1.13.1, but again this is a guess. Your best bet might be to upgrade to 1.15, as both COVID-Net and COVIDNet-CT are tested on 1.15.

from covidnet-ct.

I solved the problem by upgrading to Tensorflow 1.15. Perfect. You have done great work. Thank you.

I'm not very familiar with Python and looking for some help to make the inference request available as a local web service, to which I can communicate with JSON. Any idea?

from covidnet-ct.

I'm not too familiar with building web services, so I won't be able to help you on that end.

from covidnet-ct.

I'm closing this issue now, as the original problem was resolved as a version issue. The question about web services is best asked elsewhere, as issues should be limited to problems with the code, questions about the code/methodology, or potential enhancements.

from covidnet-ct.