This year was the 30th edition of the British Machine Vision Conference (BMVC). BMVC is one of the largest international image processing and pattern recognition organizations.

The EMBEDDED AI workshop for REAL-TIME MACHINE VISION (EMAI) focused on pattern recognition and computer vision for real-time, low-processing systems. Several papers were presented with innovations in autonomous vehicles, smart cities, industry-oriented computer vision and neural network optimization. The last theme was a major focus, since neural networks are used in visual pattern recognition and are becoming computationally expensive, a consequence of more elaborate architectures.

At this workshop, Computer Science PhD student and Smart Sense member Artur Jordão presented the paper “Pruning Deep Networks using Partial Least Squares”, which proposed a new method of reducing the computational cost of neural networks, focusing on convolutional neural networks.

In the context of visual pattern recognition, convolutional networks have been achieving state-of-the-art results in various applications, for example image classification, pose estimation, object detection and facial recognition. Moreover, in some applications convolutional networks are capable of achieving better results than humans.

In the following link you can check the publication of Paper:

https://homepages.dcc.ufmg.br/~william/papers/paper_2019_BMVCW_Jordao.pdf

Data: 10/09/2019 até 12/09/2019Local: Cardiff, Wales, UK.
Paper: Pruning Deep Networks using Partial Least Squares
Autors: Artur Jordão, Ricardo Kloss, Fernando Akio e William Robson Schwartz