Building a Pneumonia Classification Application

[JNoh01]

I am doing a project for Building a Pneumonia Classification Application.
I've downloaded the notebook files from the following link:

https://github.com/intel-iot-devkit/pneumonia-classification

When I run the code in pneumonia_classification_jupyter.ipynb, I got the following error:

Exception: Path to the model ../resources/FP32/model.xml doesn't exist or it's a directory

In the resources directory, there's no FP32 folder. How can I get model.xml file?

[JNoh01]

I raised the issue in the GitHub repo and waiting for a pull-request.  In the resources folder, there's no FP32 folder, but there's model.pb file

in the model folder.

[JesusE_Intel · Accepted solution]

Hi Jung Soon,

Looking at the repository, there is a TensorFlow format (model.pb) model in the resources/model directory. The setup.sh contains the model optimizer command to convert the model.pb to IR format that is being referenced by the Jupyter Notebook. You will need to run that script to generate the IR files.

#Optimize the model
cd /opt/intel/openvino/deployment_tools/model_optimizer/
python3 mo_tf.py -m $BASE_DIR/resources/model/model.pb --input_shape=[1,224,224,3] --data_type FP32 -o $BASE_DIR/resources/FP32 --mean_values [123.75,116.28,103.58] --scale_values [58.395,57.12,57.375]
python3 mo_tf.py -m $BASE_DIR/resources/model/model.pb --input_shape=[1,224,224,3] --data_type FP16 -o $BASE_DIR/resources/FP16 --mean_values [123.75,116.28,103.58] --scale_values [58.395,57.12,57.375]

Hope this answers your question.

Regards,

Jesus

[JNoh01]

I am working on Windows 10 environment.  So, I cannot run setup.sh as it is. Instead, I ran the following code:

!python C:/OpenVino/openvino/deployment_tools/model_optimizer/mo_tf.py -m $BASE_DIR/resources/model/model.pb --input_shape=[1,224,224,3] --data_type FP32 -o $BASE_DIR/resources/FP32 --mean_values [123.75,116.28,103.58] --scale_values [58.395,57.12,57.375]

After running the above code, I got the following result:

[ SUCCESS ] Generated IR model.
[ SUCCESS ] XML file: C:\Users\USER\Documents\pneumonia-classification-master/resources/FP32\model.xml
[ SUCCESS ] BIN file: C:\Users\USER\Documents\pneumonia-classification-master/resources/FP32\model.bin
[ SUCCESS ] Total execution time: 774.89 seconds. 

However, after running pneumonia_classification_jupyer.py, I got the following error message:
----------------------------------------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
<ipython-input-10-74cd54106bc1> in <module>
    381 
    382 if __name__ == '__main__':
--> 383     main()

<ipython-input-10-74cd54106bc1> in main()
    372                 file_name = file.split('/')[-1]
    373                 output_file = "{}/{}".format(output_dir, file_name)
--> 374                 save_img(output_file, img1, file_format='jpeg')
    375 
    376     log.info("Success")

<ipython-input-10-74cd54106bc1> in save_img(path, x, data_format, file_format, scale, **kwargs)
    120                       'RGBA images, converting to RGB.')
    121         img = img.convert('RGB')
--> 122     img.save(path, format=file_format, **kwargs)
    123 
    124 

~\Anaconda3\lib\site-packages\PIL\Image.py in save(self, fp, format, **params)
   2079                 fp = builtins.open(filename, "r+b")
   2080             else:
-> 2081                 fp = builtins.open(filename, "w+b")
   2082 
   2083         try:

FileNotFoundError: [Errno 2] No such file or directory: '../output/PNEUMONIA\\person1946_bacteria_4874.jpeg'
----------------------------------------------------------------------------------------------------------------------------
In the output folder, empty files of result.txt and stats.txt exist (No image files saved). 

[JNoh01]

BTW, I cannot upload (attach) files here.  Could you please show me another way of uploading files?

The CPU of my PC is Intel(R) Core(TM) i7-8700CPU@3.20 GHz 3.19 GHz.

The graphic card of my PC is Nvidia GeForce GTX 1660.  Is there any possibility of affecting image processing/displaying due to the Nividia graphics card?

[JesusE_Intel]

Hi Jung Soon,

The GitHub repository lists Ubuntu 16.04 as a requirement, the issue may be related to the windows path. However, I have not used this repository before, I recommend posting you issue on the GitHub Issue section.

Regards,

Jesus

[JNoh01]

Instead of running pneumonia_classification_jupyter.py on the notebook cell, I executed the following command:

!python pneumonia_classification_jupyter.py -m ../resources/FP32/model.xml -o ../output

After running the above command, got result.txt and stats.txt file, which are not empty.  Also, after copying
PNEUMONIA folder to the output directory, I also got image files, where class_activation heatmap is overlaid
over the X-ray image in case of pneumonia.  The wider the yellow color area, the higher the probability of pneumonia is. 
The file not found error is caused as there are no base X-ray image files to be overlaid. Looks like in the
pneumonia_classification_jupyter.py file, there is no code copying PNEUMONIA folder to the output directory
(I manually copied the folder).