Hi all. Newbie here. May I check how can I add training graph of accuracy, precision, mAP etc to tflite? I heard need to open a separate training console and add in the checkpoint path.
If you’re referring to checking Tensorboard while using the Object Detection Model Maker API I was discussing this a bit in this Stackoverflow question here, but never ended up implementing it.
I think you’re looking for
tensorboard --logdir
Whether that logdir is the same as the checkpoints directory in Model Maker API as it is in the regular TFOD API I’m not totally sure.
Maybe @khanhlvg might have an insight?
**The above Stackoverflow link refers to how to run a Model Maker API training session so you can export a model from any checkpoint file (it’s kinda handy to have in your bookmarks if something goes wrong etc. and you’ve still got the checkpoints folder saved).
Or are you talking about on device training?