Project_MVP.md

MVP

In our project we are trying to analyze and detect the damages in the car body using images, hoping to build a model to assist certain entity in damaged cars capturing using deep learning methods.

For a starter, we tried to use a deep learning model from amazon aws and with cnn architecture, and transfer learning vgg16 to see how it predict every image, that results in producing list of occurances of every prediction.

Figure 1

these are the common types of image predictions

Figure 2

here a close up look to the first 10 types of prediction

In further steps, we are planning to classify the images whether to be damaged or not, and try to locate the place of damages in every car.