ValueError: Dimension 0 in both shapes must be equal, but are 1 and 255. Shapes are [1,1,1024,75] and [255,1024,1,1]. for 'Assign_360' (op: 'Assign') with input shapes: [1,1,1024,75], [255,1024,1,1]. about tog HOT 7 CLOSED

Did you use the pre-trained model provided in the repository?

from tog.

Did you use the pre-trained model provided in the repository?

Thank for your reply! I have use pre-trained model. it was run successfully.but i am failed in my model.By the way, the h5 model provided by /qqwweee/keras-yolov3 is the same error(he convert .weight to .h5)

from tog.

Changes need to be made if you are using another model, especially the model with a different neural architecture. Could you provide more details on the model you are using?

from tog.

Changes need to be made if you are using another model, especially the model with a different neural architecture. Could you provide more details on the model you are using?

Thank you, this is the link
https://www.dropbox.com/home?preview=trained_weights_final.h5

from tog.

You need to share a public link. The one you provided is private and only works for you.

from tog.

sry
https://www.dropbox.com/s/dglx6rrr5vu2wjx/trained_weights_final.h5?dl=0

from tog.

Hi, I can reproduce the error you mentioned:

InvalidArgumentError: Dimension 0 in both shapes must be equal, but are 1 and 18. Shapes are [1,1,1024,75] and [18,1024,1,1]. for 'Assign_360' (op: 'Assign') with input shapes: [1,1,1024,75], [18,1024,1,1].

When you train your own model, did you modify the neural architecture? If you did, make sure you update the models/yolov3.py file accordingly. Otherwise, the default architecture provided in this repository is incompatible with the model weights.

from tog.