Project of Deep Learning
Create a Deep Learning model that is able to classify 20 IMAGES classes from the Pascal VOC 2007 dataset. The stages include: load modules and libraries, load datasets, define input sizes, load datasets, one-hot labels, view datasets, classification models, compile models, normalize data, data augmentation, custom training functions, train the model, plot training history, evaluate models, save models into .h5 files, test on Random Images from the Internet. By using the InceptionV3 model the test accuracy obtained was 68.85%. The InceptionV3 model is the best model found during experiments conducted similarly on the VGG16, VGG19, Restnet, Xception, and Alexnet models.