I am trying to train peoplenet using TLT 3.0. I am therefore using the detecnet jupyter notebook with the resnet34 pretrained model.
When I run the TAO training command, the training stops as soon as I receive the first warnings.
Could this be a problem of dataset incompatibility? since the classes in my spec file are bag, face, and person, whereas in the dataset I downloaded initially the classes were car, pedestrian and cyclist.