As announced in a previous post, we are hosting Prof. Matthieu Cord, from Pierre et Marie Curie University / Paris. The seminar last Friday was a total success and the slides are available to download here. Tomorrow he will offer us the second seminar on Deep Learning and, on August, a mini-course. The details are below.
Seminar on Friday 1st July, 17h00
Deep learning and weak supervision for image classification
Attention: the previously announced room has changed to FE-03 at FEEC/UNICAMP (pointer K). Free admittance up to the room’s capacity.
Deep learning and Convolutional Neural Networks (CNN) are state of the art for many visual recognition tasks, e.g. image classification, or object detection. To better identify or localize objects, bounding box annotations are often used. Those rich annotations are costly to get, motivating the development of Weakly Supervised Learning (WSL) models.
We discuss several strategies to automatically select relevant image regions from weak annotations (e.g. image-level labels) in deep CNN. We also introduce our architecture WELDON for WEakly supervised Learning of Deep cOnvolutional neural Networks. Our deep learning framework, leveraging recent improvements on the Multiple Instance Learning paradigm, is validated on several recognition tasks.
Mini-Course on Deep-Learning on August 9th (Tue), 11th (Thu), 12th
(Fri), 15th (Mon), 18h~20h
Provisional program: Deep-learning for Vision, From LeNet to Imagenet, Tricks for Training, Transfer Learning, Fine Tuning, Extensions and Applications (Supervised Segmentation, Automatic Captioning, Metric Learning, Generative Models).
Pre-req: Applicants are expected to know the basics about Deep Learning and Machine Learning. This will not be an 101-course: You are expect to have theoretical and practical knowledge equivalent to the 3 basic tutorials of deeplearning.net/tutorial/
Applicants should send Prof. Eduardo Valle an e-mail at email@example.com with subject “[Deep Learning Mini-Course 2016]” with:
1) Full name;
2) Current position (undergraduate, m.sc., ph.d., faculty, professional position, etc.);
3) Current affiliation;
4) Ten-line paragraph with previous work/study/experience on Machine Learning, and motivation for doing the course.