How to Train the Models

The repository will contain the code to train multiple networks on the dataset. The trainers.py file can be used for the purposes of training the available models. The default PyTorch object detector models are available for training.

Known Issues

Following are some known issues with the code:

• Negative examples do not work while training. They increase the loss to infinity from which the network never recovers.

Tested On

The code has been tested to run on:

• macOS 12.4 Monterey on Intel Macbook Pro 2019
• Google Colab
• Kaggle