Facing issue while using Peoplenet model for transfer learning with another class

[user86169]

Docker_tag:–> v3.21.08-py3
Network Type → Yolov4
Training Spec File → spec.txt.txt (2.3 KB)

Hi,

I have downloaded peoplenet pretrained ngc model named as “resnet18_peoplenet.tlt”. I want to use this model for transfer learning by freezing layer 3. I have set this pretrained model in the resume_model_path in the config file. While I am trying to train along with another class ‘phone’, facing one error i.e

( ValueError: Cannot parse the checkpoint path. Did you rename it? )

do I need to rename the file?
I have also attached the configuration file for your reference.

Looking forward for any solution from your side.

[Morganh]

The peoplenet pretrained ngc model is based on detectnet_v2 network.
So, you can only run with detectnet_v2 network instead of yolo_v4 network.

[user86169]

can I have a detectnet_v2 network architecture image.

[Morganh]

Do you mean a pretrained model trained with detectent_v2 network?
Actually peoplenet, traficcamnet, lpdnet, dashcamnet ,facenet, are trained via detectnet_v2 network.
More info are mentioned in https://docs.nvidia.com/tao/tao-toolkit/text/purpose_built_models/peoplenet.html