Identification and Classification Scrap Materials and Hazard

• 

  Graphic Table of Images Predicted

We developed this proyect based on the difficulties presented on the scrap classification and the risks than can occur when handling hazard materials

This project, implements a CNN to identify materials on scrap, and shows a table with materials identified

We built the CNN based on the tensorflow framework

We ran into multiple difficulties, the one most notable, the database, we did not have data to train the model, so we had to do a big search, and dive into photoshop to adapt and recreate scenarios to simulate the conditions of a scrap park

We are really proud of the graphic table that the prediction program outputs, this to make easier to understand what the CNN is doing and how it thinks

We learned a lot of image classification, the differents approachs, and tools to make one robust CNN

We have different imporvements, one of them implement object detection with darknet and YOLO, with this, we will have an even more user friendly system and more feedback from the system, as well as implement multiple camera feed, to be able to see on a tridimensional space, and detect the most materials possible.

Built With

• matplotlib
• photoshop
• python
• scikit-learn
• tensorflow