Download to read offline
More Related Content
Similar to OrangeNXT - High accuracy mapping from videos for efficient fiber optic cable laying
![[DSC Europe 25] Stefan_Milosevic - AI and digital twins in precision oncology...](https://cdn.slidesharecdn.com/ss_thumbnails/nnmuciuxr2ugh4d9pzkg-1-251126104228-148c7fe8-thumbnail.jpg?width=640&height=640&fit=bounds)
OrangeNXT - High accuracy mapping from videos for efficient fiber optic cable laying
- 1. High accuracy mappingfrom videos for efficient asset management in cities Jeroen Emile Delcour
- 2. Jeroen Emile Delcour •BSc. Psychobiology • MSc. Molecular Neuroscience • Data scientist
- 3. Importance of mapsand surveying •Navigation •Determine land ownership •Construction and maintenance
- 4.
- 5. Valcamonica rock art,Paspardo r. 29, topographic composition, 4th millennium BCE
- 8.
- 9. Essence of mapping 1.What? 2. Where? Camera GPS
- 10. What about aerialphotography? • Survey large areas quickly • Low resolution (~15cm per pixel) • Rarely up-to-date • Expensive • Many blind spots in cities
- 11. Drones? Way too manyno-fly zones…
- 12. IMU GPS • Cheap • On-demandand fast (thus always up-to-date) • High-resolution images • Works great in cities
- 14. GPS Images + ? We haveall the data we need
- 16. A PROPOSAL FORTHE DARTMOUTH SUMMER RESEARCH PROJECT ON ARTIFICIAL INTELLIGENCE J. McCarthy, Dartmouth College M. L. Minsky, Harvard University N. Rochester, I.B.M. Corporation C.E. Shannon, Bell Telephone Laboratories August 31, 1955 We propose that a 2 month, 10 man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire. The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer. http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html
- 17.
- 18. We need: • Data •Data • Data • Data • Data • Data • Data Convolutional neural networks
- 19. Coverage of GoogleStreet View
- 20.
- 21. Academic datasets don’tcontain relevant labels
- 22. The bigger picture GPS Images +? Municipal planning map (labeled data!) Forward pass Backpropagation
- 23.
- 25. Mapping between animage and the world
- 26. Mapping between imageand world Camera properties GPS & IMU
- 27. Orthoprojection One-to-one mapping ofpixels to geographic coordinates
- 29. Now we havelabeled data! But municipal maps are far from perfect…
- 30. The student becomesthe master Training Trained modelTraining data
- 31. The student becomesthe master Training
- 32.
- 33. Prototype map WIP Driveways andgardens (not part of municipal maps)
- 34. Turns out theworld isn’t flat Orthoprojections aren’t great for 3D objects
- 35. How do wemap objects? • Trees • Lampposts • Traffic signs • Etc.
- 36. Finding labeled imagesof trees COCO dataset •330K images (>200K labeled) •1.5 million object instances •80 object categories •91 stuff categories 🌳
- 37. Municipal maps havetree points
- 38. Mapping between imageand world • Estimate bounding box size using distance from the camera • Still far from perfect training data
- 39.
- 40. 1. Detect objectin panoramas 2. Intersect base of object with ground plane Object localization α h d
- 41.
- 42. Drive-through mapping • Fast •Cheap • Up-to-date • High resolution • Works great in cities