Silicon Valley inspired binary classifier to identify hot-dogs and not-hot-dogs. Can be used to identify anything from missiles in the air to the ones in your pants. Must be used responsibly.
After generic training and absolutely zero effort in fine tuning, the model achieved a training accuracy of 100%, validation accuracy of 73%, and testing accuracy of 57%. Please contribute for betterment.
After using transfer learning with a pre-trained model - ResNet50, the accuracy of the model is increased up to 94% for the training data and 91% for the validation data.
Pull requests welcome.
https://www.kaggle.com/datasets/dansbecker/hot-dog-not-hot-dog