Doubt in command Unit 3: Train your own TensorFlow

Course Support ROS Deep Learning With TensorFlow 101
1

Hello,

I’m following the Unit 3 and in Step 5 we call the train.py script with some flags:

  • dataset and val_dataset: are both new files
  • classes and num_classes: correpond to the new classes

But I don’t understand what’s the point of “weights” and “weights_num_classes” parameters. As far as I know, the “yolov3.tf” corresponds to another previous example, and the “num_coco_classes” is 80.

I don’t understand these two parameters. If you can give me some explaination I would be happy.

Thanks
Juan

2

Hi,

We are using the previous example DeepLearning Model Yolo with 80 clases as base for us to afterwards retrain it for only One or two clases ( in thi scase mir_robot and objects ). So we need that data to start to be able to retrain the model.

1 Like
3

Hi duckfrost,

Thank you for your answer.

Juan