Keras symbolic inputs/outputs do not implement `__len__`.

[rhea-19]

Hi,
I am getting an error while running the notebook you've uploaded. I've attached the image, kindly look into it. It would be of great help if you could help me rectify it. Thank you!

[saunack]

Hi @rhea-19
Can you please try with keras~=2.8.0? Also, the SSD model summary might not work, so you will have to comment that.
Also, please add tf.compat.v1.disable_eager_execution() after the imports. It works on Google colab that way for me.

I will add these to the README and make the necessary changes to the notebook as well meanwhile

[saunack]

Can you please check the updated code in master? commit 7f40ddfb

[rhea-19]

Hi! Thank you for the reply, I'll check it out. Also, can I get help with training the model using a non-MNIST dataset?

…

On Thu, Jul 21, 2022, 16:36 Kumar Saunack ***@***.***> wrote: Can you please check the updated code in non_trainable branch? — Reply to this email directly, view it on GitHub <#3 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AWGV7JSJVFBS7FIIO55SR6LVVEVMVANCNFSM536V7ECA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[VanqCoding]

Hi there Kumar,

Thanks for your work, gotta respect it when people implement stuff from scratch.
I am currently trying to load other data from disk to train.
How do I proceed if I have my training images in a folder and the labels as a txt file with:
"imgpath xcenter, ycenter, w, h, class"
Or how to load common train datasets properly such as Pascal VOC/KITTI.
Any help is appreciated.