Object detectors are typically learned based on fully-annotated training data with fixed pre-defined categories,In such scenario, only the original training set annotated with the old classes and some new training data labeled with the new classes are available.how to extend category?
According to your comment above, the new classes have label, right?
yeah, new training data labeled with the new classes
So, why you ask “How to extend category at Object Detector model without Relabeling?”
the old data have not new category label, the new dataset have not old category label,it can add new category on old tlt model?
OK, I am afraid you want to add new dataset with new classes. Yes, it is possible. Please modify the training spec and then run training.
how to modify?
add new target_class_mapping in the training spec and then run training with new dataset & old tlt model?
Yes, adding new target_class_mapping in the training spec.
You can also use the old tlt model as the pretrained model.
BTW, which network did you train? Detectnet_v2 or yolo_v4?
Training data contains new dataset and old dataset? or only new dataset ??
new dataset and old dataset.
The new dataset only have new classes and old dataset only have old classes,Will it affect each other？