Custom Object Detection TensorFlow Lite model

32 views

Skip to first unread message

Rajendar Uppulanche

unread,

Dec 16, 2022, 12:32:56 AM12/16/22

to Discuss

I am trying to develop an Object Detection Model which detects the trained object in the image attached below.

I have 20 different patterns of such image, which all all are unique in all four directions.

Eg :- I have 20 * 4 direction of image = 80 Unique Pattern Cards

If I have a card named 39, then I want to detect the card such that when you place D then it should view 39D or B then it should view 39B etc.

I have used labelImg to prepare the dataset with bounding boxes which is necessary to create the tflite model, which I would be using in the tensorflow android example app.

Currently I am facing difficulty in preparing the dataset to get best accuracy with precision. So I need help from you developers out there, what approach should I follow to build the model which gives me the maximum precision ?