In this project, the student will work on a machine vision system capable of identifying and tracking products in real time, in a controlled environment. The project shall employ a standard machine learning framework such as TensorFlow or Keras, and will make use of an existing product image database.
_Background_: The “AI Retailing Systems” team is working on a project that addresses some of the most pressing problems retailers are facing: lost revenues due to inability to offer 24/7 service, unnecessary operational costs and stock losses. The project won second prize in the ETH Entrepreneurship Club. The proposed solution enables retailers to open stores around the clock without infringing local laws, reduce the amount of workers neces-sary to operate the facilities and finally reduce stock losses by removing one source of the problem while improving security.
The project offers the opportunity to work in a real-life challenging computer vision project, with access to AI professionals who are already supporting us as advisors, and access to state-of-the-art hardware for modern deep learning. If successful, the project may also serve as entry point for a future full-time position.
The goal is to deliver, at the end of the thesis
- an implementation of a convolutional neural network able to automatically identify and track products
- an empirical evaluation of environmental conditions affecting the accuracy of the system
- if possible, an estimate of how the system could be improved by using a multi-camera setup
- Konrad Schindler (firstname.lastname@example.org)
- Alejandro Garcia-Santos
CLS Student Project (MPG ETH CLS)
Information, Computing and Communication Sciences