I am running the second lab with TensorFlow Hub for ssd_mobilenet_v2. When I call run_detector(detector, downloaded_image_path), it gives this error:
cannot compute __inference_signature_wrapper_42321 as input #0(zero-based) was expected to be a uint8 tensor but is a float tensor [Op:__inference_signature_wrapper_42321]
I have just run the lab and no issues where encountered, now the uint8 means it is expected to be of type INT but you have float. Have you changed anything in the lab? The only place uint8 is found is on
“image_pil = Image.fromarray(np.uint8(image)).convert(“RGB”)”, maybe you changed this one?
Now it is working thank you. I was trying with another .png file that’s why it was giving an error
Yeah png is different than jpg file, it has 4 channels.