Retraining Object detection model tutorials

General Discussion
, ,
#1

Hi all,

I was hoping to train an object detection model.

I came across TFHub and TF model Garden, where I think model garden is more suitable if I want to retrain models.

Inside the GitHub repo, there is no clearly tutorial or example on training your model, and then making predictions with it.

Can someone point me to the right place unless I am missing something?

I thought would the shear number of models available, there would be an end to end example, that shows the model in action and prediction outputs atleast.

Thanks,

#3

Hi Viraj,

There’s a couple of options to help you.
The easiest one is using Model Maker. This is not for on-device only! You can train regular TF models with it too.

On Model garden, you can follow these tutorials here: models/research/object_detection/colab_tutorials at master · tensorflow/models · GitHub
more specifically the Eager few shot training

Some more info here:

#4

It really seems a recurrent and hot topic.
I hope that we could plan to have a better integration also with TFHhub for transfer learning and finetuning with the refactoring in:

https://github.com/tensorflow/models/blob/master/official/vision/beta/MODEL_GARDEN.md#object-detection-and-instance-segmentation

1 Like
#5

I agree 100000% with you!
This kind of question help me get people’s attention to help with this!

#7

Thank you all for the quick replies !

No worries, I understand it takes time to do these things.

Will keep an eye out for advancements.

1 Like