Diamension mismatch after tensorset

[srijiths]

Hi,

I see a dimension mismatch when retrieving the image back using tensorget immediately after tensorset.

Steps to reproduce.

    new_shape = 416
    image_path = './demo_data/dog.jpg'
    pil_image = Image.open(image_path)
    numpy_img = np.array(pil_image, dtype='float32')
    print('raw image shape and dtype before pre-processing', numpy_img.shape, 
    numpy_img.dtype)
    # applied a reshaping function and expanded diamension on axis 0
    image = letter_box(numpy_img, new_shape)
    print('shape and dtype before creating Blobtensor ', image.shape, image.dtype)
    image = rai.BlobTensor.from_numpy(image)
    print('shape and dtype after creating Blobtensor :', image.shape)
    con.tensorset('image', image)
    img = con.tensorget('image', as_type=rai.BlobTensor).to_numpy()
    print('shape and dtype when retrieving using tensorget  :', img.shape, img.dtype)

Output

raw image shape and dtype before pre-processing (576, 768, 3) float32
shape and dtype before creating Blobtensor  (1, 416, 416, 3) float32
shape and dtype after creating Blobtensor : (1, 416, 416, 3)
shape and dtype when retrieving using tensorget  : (1, 1, 416, 416, 3) float32

There is a dimension mismatch when retrieving the image using tensorget immediately after tensorset.

Thanks ,

[hhsecond]

Hi @srijiths, Thanks a lot for the issue. The issue is fixed in master. We are pushing the new version to pypi soon

[srijiths]

@hhsecond Thank you Sherin for the quick turn around time..Highly appreciated..

[hhsecond]

The new version is out in pypi