Open lecture as a conclusion of the first edition of the Deep Learning for AI course of Master MET: Cristian Canton & Amaia Salvador

Website & register

DATE AND TIME

Tue, December 19, 2017    3:00 PM – 5:30 PM CET

LOCATION

Campus Nord UPC (Aula Màster A3) Jordi Girona 1-3 08034 Barcelona

Schedule

15:00 Poster session

15:20 Best project presentations

16:00 Amaia Salvador, "Pic2Recipe"

16:30 Cristian Canton, "Machine Learning at Facebook scale: challenges and examples"

17:30 End of the lecture

You are invited to prepare this lectures beforehand with the course material available online at:

http://dlai.deeplearning.barcelona/

The organisers may limit the access to the room to ticket holders only.

Contents of the talks

16:00 Amaia Salvador, "Pic2Recipe"

In this work, we introduce Recipe1M, a new large-scale, structured corpus of over 1m cooking recipes and 800k food images. As the largest publicly available collection of recipe data, Recipe1M affords the ability to train high-capacity models on aligned, multi-modal data. Using these data, we train a neural network to find a joint embedding of recipes and images that yields impressive results on an image-recipe retrieval task. Additionally, we demonstrate that regularization via the addition of a high-level classification objective both improves retrieval performance to rival that of humans and enables semantic vector arithmetic. We postulate that these embeddings will provide a basis for further exploration of the Recipe1M dataset and food and cooking in general.

Read more about pic2recipe on: MIT News, BBC, Wired, VilaWeb, La Vanguardia

Short bio

Amaia Salvador is a PhD Candidate at Universitat Politècnica de Catalunya under the advisement of Professor Xavier Giró and Professor Ferran Marqués. She obtained her B.S. in Audiovisual Systems Engineering from Universitat Politècnica de Catalunya in 2013, after completing her thesis on interactive object segmentation at the ENSEEIHT Engineering School in Toulouse. She holds a M.S. in Computer Vision from Universitat Autònoma de Barcelona. During the Summer of 2014, she joined the Insight Centre for Data Analytics in the Dublin City University, where she worked on her master thesis on visual instance retrieval. In 2015 and 2016 she was a visiting scholar at the National Institute of Informatics and the Massachusetts Institute of Technology, respectively. Her current research focuses on computer vision, multimedia retrieval and image segmentation.

16:30 Cristian Canton, "Machine Learning at Facebook scale: challenges and examples"

Facebook serves everyday 2 Billion users, helping connecting the world at an unprecedented scale. ML plays a crucial role to deliver but training and deploying algorithms require lots of engineering that are not usual to discuss in an academic environment. During this talk, we will review some of the ML challenges that Facebook deals with and some concrete examples: from AR and to detection of harmful content. Cristian will also open a Q&A with the attendees, so bring plenty of questions!

Short bio

Cristian Canton is an engineering lead at Facebook in Seattle where he currently manages the computer vision team within the objectionable content domain (i.e. detect and remove all the bad content in Facebook). In the past, he was the lead for machine learning for Augmented Reality within Facebook. From 2012-16, he was at Microsoft Research in Redmond and Cambridge (UK) where he worked on large scale Computer Vision and machine learning problems. From 2009-2012, he was the lead engineer at Vicon (Oxford), bringing CV to produce visual effects for cinema industry. He got his PhD and MS from BarcelonaTech (Spain) and his MS Thesis from EPFL (Switzerland) on computer vision topics.

Acknowledgements

This event is kindly supported by Vilynx.