![Page 1: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/1.jpg)
![Page 2: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/2.jpg)
Reconnaissance d’objetset vision artificielle
Josef Sivic ([email protected])http://www.di.ens.fr/~josef
INRIA - Equipe-projet WILLOWENS/INRIA/CNRS UMR 8548Laboratoire d’Informatique
Ecole Normale Supérieure, Paris
![Page 3: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/3.jpg)
The rest of the class today
• Overview of course assignments and the final project (J. Sivic)
• Overview of research at WILLOW and internship possibilities (I. Laptev)
![Page 4: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/4.jpg)
Course assignmentshttp://www.di.ens.fr/willow/teaching/recvis10/
![Page 5: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/5.jpg)
![Page 6: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/6.jpg)
![Page 7: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/7.jpg)
A1: Scale-invariant feature detection
![Page 8: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/8.jpg)
A2: Stitching photo mosaics
The goal of the assignment is to automatically stitch images acquired by a panning camera into a mosaic.
![Page 9: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/9.jpg)
A2: Stitching photo mosaics
1. Detect feature points (Assignment 1)2. Match feature points and estimate planar homography between images3. Warp and composite images into a common reference frame
![Page 10: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/10.jpg)
A2: Stitching photo mosaics
1. Detect feature points (Assignment 1)2. Match feature points and estimate planar homography between images3. Warp and composite images into a common reference frame
![Page 11: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/11.jpg)
A2: Stitching photo mosaics
1. Detect feature points (Assignment 1)2. Match feature points and estimate planar homography between images3. Warp and composite images into a common reference frame
![Page 12: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/12.jpg)
A2: Stitching photo mosaics
Stitch your own images!
Images courtesy: Jose Lezama
![Page 13: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/13.jpg)
A2: Stitching photo mosaics
Stitch your own images!
Images courtesy: Jose Lezama
![Page 14: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/14.jpg)
A3: Bag-of-features image classification
The goal of this assignment is to implement a category-level image classifier to perform four-class image classification into “airplanes”, “motorbikes”, “faces” and “cars” categories.
Training and test images will be provided.
The goal is to achieve at least 90% accuracy and compare performance with a baseline image representation.
Training images
Test images
![Page 15: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/15.jpg)
A4: Simple face detectorYou will develop a simple “window-scanning” face detector and analyze its performance on real images.
The detector will be based on a linear Support Vector Machine classifier.
Training and test images will be provided.
![Page 16: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/16.jpg)
![Page 17: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/17.jpg)
AssignmentsAssignments will be implemented in Matlab
We provide a comprehensive starting package including (1) detailed instructions, (2) usually some starting code and (3) datasets for each assignment.
![Page 18: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/18.jpg)
![Page 19: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/19.jpg)
![Page 20: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/20.jpg)
![Page 21: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/21.jpg)
![Page 22: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/22.jpg)
Final projects• The final project will represent 30% of the grade.
• The topics will be announced in the class and on the course page.
• Talk to us if you want to define your own project (before you start working on the project).
• Joint projects with other courses are possible, but please talk to instructors of
both courses (before you start working on the project).
• You are encouraged to work alone but you can also form groups of two or three people. Each group will submit a single final report. Reports will be in a standard conference format submission and max 3 pages long.
• Hand out date: Nov 2.
• Project presentations (~10mins per group) and reports due on: Dec 14.
![Page 23: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/23.jpg)
Final projectsTwo suggested topics building on assignments 3 and 4:
1. Image classification of 100 object classes.
![Page 24: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/24.jpg)
Final projectsTwo suggested topics building on assignments 3 and 4:
2. Object detection and localization
![Page 25: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/25.jpg)
Any questions?
![Page 26: Reconnaissance d’objets et vision artificielle](https://reader035.vdocument.in/reader035/viewer/2022062813/56816647550346895dd9be2b/html5/thumbnails/26.jpg)