search

tensorflow / models

Models and examples built with TensorFlow
Other
76.94k stars 45.8k forks source link

How to to train custom ssd mobilenet from scratch #8339

Open yieniggu opened 4 years ago

yieniggu commented 4 years ago

How can I retrain a ssd-mobilenet-v2 from the tensorflow object detection model zoo without transfer learning. I mean every weight and not just the last layer.

Do I have to build the network architecture and the script for training or can I make some minor change to the .config file or the train.py script specified in the docs on training a custom object detector.

Will this approach improve/worsen the network accuracy/loss?

Thanks in advance for any help.

Wheest commented 3 years ago

Also looking for a way to do this, for any model using the object detection model zoo pipeline.

I seriously doubt it will improve accuracy for my use-case, but I need to train a reinitialised model from scratch to compare against.

Is there a way, following the steps outlined in the Object Detection API tutorial, to re-initialise the parameters of the model and train from scratch?

Annieliaquat commented 10 months ago

How can I retrain a ssd-mobilenet-v2 from the tensorflow object detection model zoo without transfer learning. I mean every weight and not just the last layer.

Do I have to build the network architecture and the script for training or can I make some minor change to the .config file or the train.py script specified in the docs on training a custom object detector.

Will this approach improve/worsen the network accuracy/loss?

Thanks in advance for any help.

Did you find the answer? if you could help me in this regards?