Use Anaconda to setup environment
Run this command:
conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia
Setup other package along the way.
After activate the environment. Run Jupyter Notebook by using anaconda navigate.
Unzip the data set to train the model.
At the train(model):
you can change the:
num_of_epochs = 300
to 400 or more to become more accurate.
Then change the load model here:
model.load_state_dict(torch.load("models/model_ep299.pth"))
Resize the image to 256 pixel
https://blog.paperspace.com/object-localization-using-pytorch-1/