Make a constant counter for a custom object

Hello I would like to build a system that will count custom objects passing through a conveyor belt. We can say that will be passing several objects at the same time there will be something like accumulation of objects, this system must work non-stop.

With that I would like to know which framework you guys recommend me for this application, which will be better for training and if jetson nano can meet these needs?


You can start from the below sample, which is for counting people with Deepstream SDK:



I started to use and study the deepstream/transfer learning toolkit I would like to know which training model you think is best in my situation I will not use people but use objects that were too close to each other, as I see I can’t find a good model for the application of detecting multiple objects close to each other.