Hello community!
First of all I would like to congratulate the author of the code, it's an amazing job. Very simple to reproduce and easy to understand.
I would like to ask a question. Right now I have the network trained with the GEN1 dataset. However, I have a custom dataset of 3 classes. Is there a possibility of fine-tuning the trained network with 2 classes (cars and pedestrians) to the custom dataset I have?
The idea would be to freeze the backbone (vgg-11) and train only the final layer (SSD Head).
Hello community! First of all I would like to congratulate the author of the code, it's an amazing job. Very simple to reproduce and easy to understand.
I would like to ask a question. Right now I have the network trained with the GEN1 dataset. However, I have a custom dataset of 3 classes. Is there a possibility of fine-tuning the trained network with 2 classes (cars and pedestrians) to the custom dataset I have?
The idea would be to freeze the backbone (vgg-11) and train only the final layer (SSD Head).
Best regards and thank you very much