Slimnet pre-trained model questions

[ari-ruokamo]

Hi! And thanks for your amazing source of DL information. I'm exploring the repository and the pre-trained 'slim_reshape.tflite' model.

In my python script I'm feeding input of 512x512x3 RGB-frame but I can't seem to get any sensible result out from the model. I'm suspecting that my input tensor data array is ill-formatted, or I'm interpreting the output tensor wrong (alpha mask?).

Thanks for any tips!

[anilsathyan7]

The exported model had some issue, it seems(orginal hdf5 is correct). Now it has been fixed and you can try the updated versions of the model from the same folder. A slightly different version of the model(output shape) is available at: https://github.com/anilsathyan7/Portrait-Segmentation/blob/master/mediapipe_slimnet/portrait_segmentation.tflite

[ari-ruokamo]

Edited:

Thanks for your response! After staring at my code long enough I realised I was using cv2.resize vs. I should've done reshaping with numpy. :-)

while True:

    # Read webcam frame 
    success, frame = cap.read()
    image = center_crop_resize(frame, side=512)
    image = np.float32(image)/255

    expimg = image[np.newaxis, :, :, :]

    # Invoke interpreter for inference
    interpreter.set_tensor(input_details[0]['index'], np.array(expimg, dtype=np.float32))
    interpreter.invoke()
    output = interpreter.get_tensor(output_details[0]['index'])

    # Reshape byte buffer & display the mask
    output = np.uint8(output*255.0)
    output = output.reshape(512,512,1)
    cv2.imshow('Portrait Mask', output)

[ari-ruokamo]

Issue author error, model works.