I am using tensorflow object detection.

I can use a model from the tensorflow model zoo to perform inference, and it detects several real world objects. However when I train the same model over new classes, then it is only able to identify the new classes. How to retain the learnings from the original model when it is used as base model for training against new data?


1 Answer 1


When you use new data, the weights of the notwork change according to your new data to learn recognizing them , if you want to keep the weights related to old data as well, you should retrain the whole network with all of the data meaning new + old data.

  • $\begingroup$ 1) Where do I get the data for the model zoo? 2) Does this mean building model from scratch or only re-training with entire dataset? $\endgroup$
    – variable
    Commented Aug 8, 2022 at 4:53
  • $\begingroup$ 1) if you name your data, I can help you find a good dataset. 2) you can do both, training model with new data, either with a pretrained network or from scratch $\endgroup$ Commented Aug 8, 2022 at 5:34
  • $\begingroup$ 1) General object recognition - for example the base model I currently use is SSD ResNet50 V1 FPN 640x640 (RetinaNet50). 2) How to do the training of the model on new data such that it retains old data? $\endgroup$
    – variable
    Commented Aug 8, 2022 at 7:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.