actually it is not new issue , it is specific question on object detection .

[ozzcet]

trafficstars

Hello , I am glad to meet you,

after watching your videos on object detection, I have dived in to dreams. And ı have gotten some specific questions.

Let me ask and explain ,

For example , We have more than 5 object in the images like A,B,C,D,E and we are creating more than 100 image for each of them. After that we create train and test file which is randomly selected from the source images for example %30 test %70 train. until now is ok .

after training we are going to create tf.lite and txt file what contain label name of 5 object for raspberry application for example cool. and the application works well to detect the 5 object which is given above as an example.

after many times we get another object which is "F" it is new one. And we are create a bunch of image of F as a hundred image.

so how could we add the last object in to the fist five object to work together?

I am curious about it. Should we train whole dataset after adding every new object from scratch ? Or how or what?

Best regards.