Class logistics Object recognition and Computer Vision 2020 - PowerPoint PPT Presentation

Object recognition and Computer Vision 2020 http://www.di.ens.fr/willow/teaching/recvis Class logistics

Object recognition and Computer Vision 2020 http://www.di.ens.fr/willow/teaching/recvis Lectures: Jean Cordelia Mathieu Ivan Josef Armand Gul Schmid Ponce Aubry Laptev Sivic Joulin Varol Teaching assistants: Yann Robin Labbé Strudel

RecVis20 Schedule Tuesday 16:15-19:15 Location: Salle Jaurès, 29 rue d’Ulm, Paris

Object recognition and computer vision 2020 Class webpage: http://www.di.ens.fr/willow/teaching/recvis Grading: Policy • 3 programming assignments (50%) • Instance-level recognition Assignments are • Neural networks strictly individual • Image classification competition Copy-paste of the 0p. code, results, parts • Final project (50%) of the report More independent work, resulting in a FPs can be done in groups of max 2 report and a class presentation. people

Assignment I: Instance level recognition • Part I: Sparse features for matching specific objects in images ○ Feature detector and descriptor ○ Robust match filtering techniques • Part II: Compact descriptors for image retrieval

Assignment II: Neural networks • Part 1: Neural Network’s theory: Forward pass, Backward pass − Parameter update − • Part 2: Building blocks of convolutional neural networks • Part 3: Training a CNN on CIFAR-10 dataset with PyTorch

Assignment III: Image Classification Competition • Bird image classification • Class Kaggle competition

Final project • Can be done individually or as a group of max 2 people • The proposed project topics are from the recent top-conference publications in computer vision, see example topics from 2019 here: http://www.di.ens.fr/willow/teaching/recvis19orig • Student-defined projects are welcome • Final project can be joint with another MVA course • We plan to arrange $100 Google Cloud credits for the project ○ This will be announced through Moodle before projects start • Select the topic + write project proposal • Present the work in the class • Write project report

Practical: Class Moodle • Sign up for the course with your school account via https://moodle.ens.psl.eu/course/view.php?id=1068 • In case of problems, ask for help from the Moodle Administrator

Practical: Class Moodle Submission of assignments

Practical: Class Moodle

Practical: Class Moodle

Practical: Lectures • Physical: Time: Tuesdays 16:15-19:15 Location: Salle Jaurès, 29 rue d’Ulm, Paris Sign up to each lecture by filling the form: • Zoom: The link will be announced on Moodle on the day of the lecture

Practical: Python tutorial Fill-in the Python tutorial participation form linked from the class webpage by Tue Oct 6. The tutorial will be at: INRIA/Willow, 2 Rue Simone IFF, Paris Who should participate? - Students with no or limited experience with Python. Topics covered: - Installing Anaconda. - Brief introduction to Python. - Introduction to Numpy, PyTorch for computer vision. - Using Jupyter notebooks.

Recap: 1. Register on the class Moodle 2. Fill-in Physical lecture participation form (by Mon each week) 3. Fill-in Python tutorial participation form (by Tue Oct 6) 4. Assignment submissions, discussions and announcements (e.g. lecture Zoom links) will be done on Moodle. 5. Assignment 1 – Instance-level recognition Due on Nov 3 2019

Introduction to computer vision http://imagine.enpc.fr/~aubrym/lectures/introvis20 Tuesdays 9:00 - 12:00 at Salle E.Noether (ex UV), ENS ULM. Taught by Mathieu Aubry. M1 course Covers the basics of computer vision in detail. Mathieu Aubry

Research WILLOW (J. Ponce, I. Laptev, J. Sivic, C. Schmid) is active in computer vision, visual recognition and robotics research. http://www.di.ens.fr/willow/ with close links to SIERRA – machine learning (F. Bach) http://www.di.ens.fr/sierra/ There will be master internships available. Talk to us if you are interested!