| KEYWORD |
Microelectronics
Real-Time 3D Object Pose Estimator using AI on Edge
Thesis in external company
keywords IMAGE ANALYSIS, MACHINE LEARNING
Reference persons LUCIANO LAVAGNO
External reference persons Marcello Babbi, Reply Torino
Research Groups Microelectronics
Thesis type APPLIED RESEARCH
Description The accurate and real-time estimation of 3D poses of objects has many practical applications in various
domains, such as robotics, smart agriculture and augmented reality. This thesis proposes the development of
a state-of-the-art 3D pose estimator using AI and deploying it on edge devices. The system will use computer
vision and deep learning techniques to analyze 3D point clouds and estimate the pose of specified objects,
enabling accurate and efficient object tracking and recognition. The thesis will focus on implementing the
latest deep neural network models for 3D pose estimation, data processing, and feature engineering. The
system will be designed to run on edge devices, enabling real-time processing of images and immediate
feedback to the user.
Required skills Python programming, some knowledge of Machine Learning, image processing
Notes The student will be involved in:
❑ State-of-the-art literature review
❑ SW requirements definition for edge deployment
❑ Data pre-processing and feature engineering
❑ Developing and implementing a deep learning-based model for 3D point cloud analysis
❑ HW computational requirements trade-off
Deadline 11/10/2024
PROPONI LA TUA CANDIDATURA