I’m trying to create a simple Snap effect that detects eggs. I am using the custom trained tool and have created >100 images with annotations (all my images are from Google API Open Images Dataset).
It seems like it would be pretty straightforward. However, when I train the model and download the .pb and Lens Studio files, the lens seems to be detecting eggs from all over the place
I’ve tried increasing the Confident Threshold, but the bounding box still seems to be detecting elements from off-screen and not actually around the eggs themselves.
Is there something wrong with the training data? Or do I need to use an older version of Lens Studio (it asks me to update to the latest when I open the fritz project file)?
Any help would be appreciated!