Transfer Learning Issue from Hello AI World tutorial

I’ve been successful in following the Hello AI World tutorial without error. However, after I train the model with my own dataset and labels, it says that any video input is always 100% the trained object. Even backgrounds or my hand over the camera lens.

I’m following the steps from here:
https://github.com/dusty-nv/jetson-inference/blob/master/docs/pytorch-collect.md

I thought that maybe I hadn’t taken enough sample photos, so I re-trained the model with double the photos and on a blank background to reduce complexity. But the result was the same.

Any advice? Thanks in advance.

Hi,

Could you share one of your database with us to check?
Since the issue is related to the data you used.

Thanks.

I can’t share the data outside of our organization. But your comment gives me hope that this issue is related to data input. I will perform another test with more images, or perhaps use an image set that is known to work well. Thanks!