How to continue training from previously trained model

[luvvien]

Have I written custom code (as opposed to using a stock example script provided in MediaPipe)

No

OS Platform and Distribution

Mac OS

Python Version

3.11.11

MediaPipe Model Maker version

lastest

Task name (e.g. Image classification, Gesture recognition etc.)

Object detect

Describe the actual behavior

No

Describe the expected behaviour

No

Standalone code/steps you may have used to try to get what you need

No

Other info / Complete Logs

As-is, let say i have a model that already traind on 100 images. Next day, if i have 100 more, i need to retrain all over again. if the model is able train from the previously trained model, then it would reduce the loads of hardware.

[kuaashish]

Hi @luvvien,

This feature is not currently supported by our APIs. You can consider custom implementation using the Object Detector model's save_checkpoint and load_checkpoint methods.

Thank you!!