On this mission, I convert my FruitsClassifier from Pytorch to Pytorch Lightning. I make use of a callback module, a checkpoint mode to retailer and save the mannequin state_dict. I conduct resume coaching, plot loss and accuracy from earlier trains. I test the accuracy for every class and make predictions on pictures from exterior the dataset.
Β
βοΈ Documentation:
– Convert my Pytorch model to Pytorch Lightning
– FruitsClassifier on Fruits 360
– Full Coaching on Kaggle | GoogleNet | 94% Acc
Β
π₯ Coaching Epochs
Β
π Visualize Loss and Accuracy
Β
π Visualize mannequin efficiency on check dataset
Β
π Accuracy of every class
Β
π Picture Predictions from exterior the Dataset
Β
π Accuracy of every class: inexperienced(right predictions), pink(incorrect predictions)