0

Hi there I'm quite new to this sort of stuff,

I'm at the final stages of completing my facial verification software and everything has run smoothly right up to the last section.

I've had an error.

It says that the error comes from the following and produces the following:

InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-148-2d15a2a2f252> in <module>
     11         cv2.imwrite(os.path.join('application_data', 'input_image', 'input_image.jpg'), frame)
     12         # Run verification
---> 13         results, verified = verify(model, 0.5, 0.5)
     14         print(verified)
     15 

<ipython-input-147-a92e41905a3e> in verify(model, detection_threshold, verification_threshold)
      4     for image in os.listdir(os.path.join('application_data', 'verification_images')):
      5         input_img = preprocess(os.path.join('application_data', 'input_image', 'input_image.jpg'))
----> 6         validation_img = preprocess(os.path.join('application_data', 'verification_images', image))
      7 
      8         # Make Predictions

<ipython-input-146-c90c40e2cf87> in preprocess(file_path)
      4     byte_img = tf.io.read_file(file_path)
      5     # Load in the image
----> 6     img = tf.io.decode_jpeg(byte_img)
      7 
      8     # Preprocessing steps - resizing the image to be 100x100x3

~/opt/anaconda3/lib/python3.8/site-packages/tensorflow/python/ops/gen_image_ops.py in decode_jpeg(contents, channels, ratio, fancy_upscaling, try_recover_truncated, acceptable_fraction, dct_method, name)
   1166       pass
   1167     try:
-> 1168       return decode_jpeg_eager_fallback(
   1169           contents, channels=channels, ratio=ratio,
   1170           fancy_upscaling=fancy_upscaling,

~/opt/anaconda3/lib/python3.8/site-packages/tensorflow/python/ops/gen_image_ops.py in decode_jpeg_eager_fallback(contents, channels, ratio, fancy_upscaling, try_recover_truncated, acceptable_fraction, dct_method, name, ctx)
   1240   fancy_upscaling, "try_recover_truncated", try_recover_truncated,
   1241   "acceptable_fraction", acceptable_fraction, "dct_method", dct_method)
-> 1242   _result = _execute.execute(b"DecodeJpeg", 1, inputs=_inputs_flat,
   1243                              attrs=_attrs, ctx=ctx, name=name)
   1244   if _execute.must_record_gradient():

~/opt/anaconda3/lib/python3.8/site-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     57   try:
     58     ctx.ensure_initialized()
---> 59     tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
     60                                         inputs, attrs, num_outputs)
     61   except core._NotOkStatusException as e:

InvalidArgumentError: Unknown image file format. One of JPEG, PNG, GIF, BMP required. [Op:DecodeJpeg]

However, I have checked all the images involved and they are all jpegs/jpg. I don't understand where I have gone wrong. Below is the function for verifying the images:

def verify(model, detection_threshold, verification_threshold):
    # Build results array
    results = []
    for image in os.listdir(os.path.join('application_data', 'verification_images')):
        input_img = preprocess(os.path.join('application_data', 'input_image', 'input_image.jpg'))
        validation_img = preprocess(os.path.join('application_data', 'verification_images', image))
        
        # Make Predictions 
        result = model.predict(list(np.expand_dims([input_img, validation_img], axis=1)))
        results.append(result)
    
    # Detection Threshold: Metric above which a prediciton is considered positive 
    detection = np.sum(np.array(results) > detection_threshold)
    
    # Verification Threshold: Proportion of positive predictions / total positive samples 
    verification = detection / len(os.listdir(os.path.join('application_data', 'verification_images'))) 
    verified = verification > verification_threshold
    
    return results, verified

And this is the opencv code that provides me with the errors:

capture = cv2.VideoCapture(0)
while capture.isOpened():
    ret, frame = capture.read()
    frame = frame[200:200+250,500:500+250, :]
    
    cv2.imshow('Verification', frame)
    
    # Verification trigger
    if cv2.waitKey(10) & 0xFF == ord('v'):
        # Save input image to application_data/input_image folder 
        cv2.imwrite(os.path.join('application_data', 'input_image', 'input_image.jpg'), frame)
        # Run verification
        results, verified = verify(model, 0.5, 0.5)
        print(verified)
    
    if cv2.waitKey(10) & 0xFF == ord('q'):
        break
capture.release()
cv2.destroyAllWindows()
Freddie
  • 1
  • What does `preprocess` function do? `image` variable is path of image, so if you do not read it as an image in `preprocess` function, it could be problematic. – Mohammad Khoshbin Jan 01 '22 at 20:11
  • Hi thanks for the reply. The preprocess function it reads and resizes the images file paths between 0 and 1 and then returns the image. It produces a NumPy equivalent of our image. In the error, you can see I wrote a line saying tf.io.read_file and passing through the file path. Is this wrong? If so how else could I read the image correctly? – Freddie Jan 04 '22 at 14:58
  • Hi. It is not wrong. Maybe you have some non image files or some folders in your directory, so `decode_jpeg` can not handle it. – Mohammad Khoshbin Jan 06 '22 at 10:17

0 Answers0