Overview
Explore deep learning techniques for 3D object recognition in this EuroPython 2017 conference talk. Learn how to achieve 80% accuracy across 40 categories in just three months. Discover strategies for gathering information, selecting frameworks, and acquiring data from public sources or creating your own. Follow a step-by-step approach, starting with small datasets and gradually scaling up. Gain insights on prioritizing tasks, optimizing processes, and leveraging GPU and CPU resources. Delve into parameter tuning, model versatility, and data augmentation techniques. Understand the importance of creating minimum viable products and utilizing Docker for deployment. Perfect for those looking to apply deep learning to 3D modeling, this talk provides practical guidance and real-world examples to help you overcome challenges and achieve impressive results in the field of 3D object recognition.
Syllabus
Intro
Strategy
Boxnet
Improving technique
Documentation
Optimization
Results
Conclusion
Demo
Taught by
EuroPython Conference