Train existing object detection nanodet model

[iamharshthakkar]

I am using nanodet as an object detection library in my Android (java) application, here is an example of an Android app: https://github.com/RangiLyu/nanodet/tree/main/demo_android_ncnn I also tried this Android nanodet project: https://github.com/nihui/ncnn-android-nanodet but was unable to figure out how can I achieve the end result.

My goal is to train and build a new data model on top of the existing one: by taking single image, and providing label name; I want to send these data to a Python server, which is expected to respond .bin and .param file (NCNN type)

I want to add new labels (class_names) included into my existing Android app dynamically by taking new .bin and .param files into consideration.

My question is: how should I add these new labels (class_names) into my Android app dynamically without rebuilding the application or modifying the code?

Your help will be much appreciated! Thanks :)