Open dhgokul opened 6 years ago
@dhgokul - there is currently no way of doing this. I'll leave this open for now as a feature request.
Are you adding any new classes with these new images? Or are all images for classes that were already present in the data when the model was trained the first time?
@TobyRoseman Thanks for your response . I have add new images with same classes and also add images with different classes. Is there any possible ways to retain existing model with new dataset for object detection ?
Sorry, no. This is not something we currently support for any of our models.
@dhgokul The most reliable way to do this is to train models with all the data as you collect more data. It gives you reliable, reproducible, and debuggable. You will also have more control over what data to use and in what capacity.
@srikris Thanks ! Is there is checkpoint option for object detection ? When I check turicreate API, it's available for regression create function .
@srikris Any updates on this request?
@srikris I have trained 1000 image data-set and created coreML model. Again i added 200 images to previous data-set Any possible ways to train previous coreML model with new data-set instead of train 1200 images from scratch ?