Image Classification Task thanks to Convolutional Neural Networks and Transfer Learning. In the first part the implemented model is a LeNet CNN The transfer learning part is done thanks to pretrained model (InceptionV3) over ImageNet Dataset (http://www.image-net.org).
The goal is to use two different tools to achieve the same goal, estimating which tool was better (CNN vs Transfer Learning). The same image dataset was used for both experiments.
The approach for both experiments is described into the report
For both experiments I got very good results as following
The final conclusion was that transfer learning works better for these types of tasks and that thanks to this it was possible to achieve very high accuracy (solving some problems that were instead encountered with the previous cnn).