fsns_demo.py reshape error

[Averu-zz]

Hello. I'm getting error when trying to run fsns_demo.py.

Traceback (most recent call last):
File "fsns_demo.py", line 153, in
predictions, crops, grids = network(image[xp.newaxis, ...])
File "/home/ml/Downloads/see/chainer/models/fsns.py", line 514, in call
images = F.reshape(images, (batch_size, num_channels, height, 4, -1))
File "/home/ml/env/lib/python3.5/site-packages/chainer/functions/array/reshape.py", line 98, in reshape
y, = Reshape(shape).apply((x,))
File "/home/ml/env/lib/python3.5/site-packages/chainer/function_node.py", line 230, in apply
self._check_data_type_forward(in_data)
File "/home/ml/env/lib/python3.5/site-packages/chainer/function_node.py", line 298, in _check_data_type_forward
self.check_type_forward(in_type)
File "/home/ml/env/lib/python3.5/site-packages/chainer/functions/array/reshape.py", line 40, in check_type_forward
type_check.prod(x_type.shape) % size_var == 0)
File "/home/ml/env/lib/python3.5/site-packages/chainer/utils/type_check.py", line 524, in expect
expr.expect()
File "/home/ml/env/lib/python3.5/site-packages/chainer/utils/type_check.py", line 482, in expect
'{0} {1} {2}'.format(left, self.inv, right))
chainer.utils.type_check.InvalidType:
Invalid operation is performed in: Reshape (Forward)

Expect: prod(in_types[0].shape) % known_size(=1800) == 0
Actual: 900 != 0

Any Ideas how to fix this? I have tried different shaped and sized pictures but no avail.

[Bartzi]

How did you run the fsns_demo.py script?

[Averu-zz]

with command
python fsns_demo.py /home/ml/Downloads/see/chainer/models model_35000.npz /home/ml/Downloads/see/chainer/models/vk.jpg /home/ml/Downloads/see/datasets/fsns/fsns_char_map.json

[Bartzi]

could you show me this image? /home/ml/Downloads/see/chainer/models/vk.jpg

[Averu-zz]

sure here is a link: https://imgur.com/a/sDiJdaL
I tried pictures of varying format and sizes. Getting the same error, but values of known size and actual change.

[Averu-zz]

I did not copy all files fsns_model/model to see/chainer/models. Just model_35000.npz and log files. The fsns.py and train_fsns.py wher not copied. see/chainer/models already contained a fsns.py file and I dont think I'll need train_fsfs.py to test demo. Should i replace the fsns.py?

[Bartzi]

I tried pictures of varying format and sizes

The FSNS Model is trained on pictures of size 600x150. Furthermore, the FSNS dataset consists of images that always show four independent views of a street sign. So the model won't work on an image like the one you provided.