Face detection and recognition was the challenge made by the VISUM Summer School 2017 to 18 international teams of researchers. Fraunhofer Portugal AICOS took the second place with an outstanding presentation, having received an Honourable Mention. The event occurred from 7th to 14th of July, in Porto.
Carina Figueira, Duarte Folgado and João Costa are the three researchers from Fraunhofer Portugal AICOS (FhP-AICOS) that participated in this summer school as an opportunity to develop several competences, mainly centred on Machine Learning and Computer Vision, like Motion Analysis and Facial, Emotion and Activity Recognition. These skills were tested in a final competition, attended by 18 international teams.
Activity recognition, facial recognition and emotion analysis were the three challenges for this summer school. In response, the FhP-AICOS’ team used Convolutional Neural Networks methods. In some cases, they built their own network architecture and, in other cases, leveraged techniques such as fine tuning, in order to adapt well proven networks to the specific tasks at hand. Clever use of pre-processing techniques and solid Machine Learning development methodologies were also required to achieve the final score in the competition.
Machine Learning is the main research topic chosen by Carina Figueira and this summer school was a complement of her work: ‘VISUM allowed me to improve my knowledge in this area. I fully recommend it!’. For Duarte Folgado, ‘It was a terrific experience’, that introduced him to Machine Learning and Computer Vision. On the other hand, João Costa sees VISUM as an opportunity to learn advanced Machine Learning concepts, having highlighted the ‘excellent environment for sharing knowledge and debating ideas in cutting edge aspects of this field. The competition was particularly interesting for implementing state of the art approaches in complex tasks’.
Despite the high competition standard, the FhP-AICOS’ teamwork overcome themselves and showed the value of the centre’s researchers, taking into account all the developed projects, master thesis and published papers which are based in these techniques. FhP-AICOS can now apply Machine Learning in even more technologies than the ones currently in use.
This was the 5th edition of VISUM 2017 (VISion Understanding and Machine intelligence), organized by INESC TEC. Every year, the summer school gathers PhD candidates, Post-Doctoral scholars and researchers from academia and industry with research interests in Computer Vision and Machine Intelligence. The final competition was based on the topics presented during VISUM: theoretical, “hands-on” and industrial sessions. These topics comprised Computer Vision, Machine Learning, Computer Graphics and Deep Learning.