Data&AI in TomTom: How are are transforming mobility
summary
TomTom explains that GPS and sensor fleets produce 18,000 times around the world per day and uses SVMs, random forests, convolutional neural networks and ensemble/stacking to classify road features. A GAN variant improves lane segmentation versus plain CNNs, and domain adaptation handles brightness and country shifts. HD maps and a live in-car format are delivered; a Road Trips recommender moves from ~40% like-probability to ~80% within five recommendations and ~87% within ten.
AI-written summary · about
chapters (19)
- 0:00Start
- 1:46Solution: Flying Cars?
- 2:06What's the Recipe?
- 3:59Types of Analytics
- 6:52Software Development vs Machine Learning
- 8:32Machine Learning Applied: Apple vs Bananas
- 9:14Linear Classifier
- 9:29The World is Not Linear
- 10:01Ensemble Learning
- 11:23Stacked Learning
- 12:03Artificial Neuron
- 12:28Artificial Neural Network
- 13:08Deep Architectures: ResNet
- 13:38Deeper is (almost) better
- 14:59Generative Adversarial Networks
- 16:55EL-GAN: General Framework
- 18:12Dynamic Adaptation
- 21:01From Personal to Personalized
- 21:56Preliminary Results
related talks
19mHow we leverage AI to build great maps at massive scale
19mAI Trends in Intelligent Transport Systems
19mChallenges & Opportunities for Building Crowdsourced Mapping Services for Autonomous Driving
31mDeep learning for self-driving vehicle technology
41mSpotlight on AI use cases and innovators around the world
45mArtificial intelligence: How data is evolving the future of technology
46mML Use Cases at Think Big Analytics
04mInterview