nn_cifar10

hello, i am trying the neural network code of cifar-10 which actually recognizes objects like cat,dog,vehicle etc. But i faced some bugs here.Which i am unable to solve…Please help me, i have attached the bugs and the code.
nn_cifar10.py (1.36 KB)



Looks like the file is corrupted, the network is only 88KBs. Please download it again using this github link:

https://github.com/openmv/openmv/blob/master/ml/cmsisnn/models/cifar10/cifar10.network?raw=true

Copy the file to the camera, and make sure it’s written (safe remove the disk).

Sir, can we use this nn for image recognition? If yes, then how we can prepare a dataset with images of some people in a .network format?

https://github.com/openmv/openmv/blob/master/ml/cmsisnn/README.md

hello sir,Thank you for your help,but an error came out.Screenshot attached…
Thanking you,
Elight-spm

It means the network is too big. So, you need to reduce the net size. The M7 doesn’t really have enough heap ram to run larger nets.

I don’t think cifar10 works on the F7, try cifar10_fast

Sir, how can I do this?

Okay sir, I am trying…

Hello sir,
Can I make a custom dataset in .network extension with images…?

Yes, please read the guide linked to above in a previous post. It’s quite detailed.

Hi, I am having the same problem with Cifar10. I would like to try the suggestions given but the link to GitHub appears to be broken. Could you kindly fix it?
Thank in advance

It’s been removed. We are all TensorFlow now. Please use Edge Impulse to Train a CNN for the OpenMV Cam.