Training ssd-mobilenet on jetson nano from scratch

iam new to jetson nano and deep learning, i followed hello AI world tutorial and everything works fine.
but iam facing a problem with training my own ssd-mobilenet, iam not finding any clear tutorial on training from scratch, like labeling type (pascal voc, kitti,…), or how to actually train the model.
i worked with yolov3 on colab and it was easy.
any help?


Have you checked this tutorial before?

This tutorial can guide you to retrain a custom model with ssd_mobilenet.


thank you for the info, do you have any idea about the annotation format of the dataset? is it pascal voc?

It supports OpenImages dataset format and Pascal VOC dataset format. For custom datasets, Pascal VOC format is used.