Full-day BME TMIT & NVidia DLI workshop: Fundamentals of Deep Learning for Computer Vision

In December 2019, we successfully held a full-day NVidia Deep Learning Institute (DLI) workshop on Fundamentals of Deep Learning for Computer Vision in Budapest, Hungary. In this beginner-level workshop, the participants learn image classification and object detection techniques, furthermore, some important aspect of deep learning-based modelling and performance improvements are discussed as well.

The NVIDIA Deep Learning Institute (DLI) trains developers, data scientists, and researchers on how to use deep learning and accelerated computing to solve real-world problems across a wide range of domains. With access to GPU-accelerated workstations in the cloud, the participants learn how to train, optimize, and deploy neural networks using the latest deep learning tools, frameworks, and SDKs.

The DLI workshop was started a little bit after 9 am — as the more than 50 participants arrived.

First, the fundamentals of deep learning were briefly introduced. Deep neural networks are nowadays the most important technique of machine learning and artificial intelligence. Deep learning models are capable of representation learning and modelling jointly, which can outperform other machine learning methods with feature engineering, in case of a large amount of data.

After the brief introduction of deep learning, the attendees started to work on the NVidia DIGITS platform that can be used to rapidly train the highly accurate deep neural network for image classification, segmentation and object detection tasks. Around 10 am most of the participants already trained their first deep neural net. We continued our work with modifying the neural network architecture, training and deploying the model in the DIGITS and Python environments.

We had delicious pizza for lunch, and after some pleasant chat, we continued the workshop. At the second part of the day, the attendees had to apply modifications to make the neural networks detect objects on images. First, we did it by brute force, then by introducing neural network architectures, specially designed for object detection tasks.

At the end of the day, the attendees had to take an assessment. Those, who completed the assessment, received an NVidia DLI certificate.

The workshop was led by Dr. Bálint Gyires-Tóth (BME TMIT, NVidia DLI Certified Instructor and University Ambassador); Gergely Hajgató and Hamdi Abed, deep learning PhD students, helped as teaching assistants.

Our host was EIT Digital, and the venue was the EIT Digital Co-Location Center Budapest. EIT Digital boosted the creativity of the participants with superb coffee and snacks, while the soft drinks and pizzas for lunch were sponsored by NVidia.

The next NVidia DLI workshop in Hungary will be in Q2 or Q3 of 2020. Keep updated in the following Twitter channel: https://twitter.com/bapalto