Breaking speed records. Sharpening the grainiest of images. Delivering candy with autonomous robots.
It’s been an eventful week at the International Conference on Machine Learning in Stockholm, Sweden, as NVIDIA put a spotlight on the future of AI and deep learning.
Attendees flocked to our booth, packed presentations of the latest breakthroughs from our research team and partners, and joined us in spotting the bot.
An AI to Fix Your Photos
Our research team presented their breakthrough deep learning-based technique that fixes grainy photos simply by looking at corrupted examples.
The team used NVIDIA Tesla P100 GPUs with the cuDNN-accelerated TensorFlow deep learning framework to train their system on 50,000 images in the ImageNet validation set.
This approach could one day help people fix their favorite holidays snaps, but there are also bigger implications. In healthcare, for example, the technique could be applied to enhance MRI images.
Breaking Speed Records
We’re speeding ahead to accelerate autonomous vehicle safety.
During ICML, we announced that NVIDIA’s Applied Deep Learning Research team has taken over the No. 1 spot for per-pixel semantic segmentation on the Cityscapes computer vision benchmark.
Cityscapes was created by Daimler R&D and Bosch to help autonomous vehicles better understand urban street scenes. It’s comprised of 20,500 images from 50 European cities, gathered during different seasons and weather conditions.
A computer vision system’s ability to understand its surroundings is key to getting an autonomous vehicle and its occupants safely to their destination. This understanding comes from being able to quickly and correctly identify nearby objects, and determining how to navigate them safely.
The team’s Cityscapes performance gains are part of NVIDIA’s commitment to safety in autonomous driving.
Smorgasbord of AI Research
Research partners from our NVIDIA AI Labs initiative presented their recent breakthroughs, powered by GPUs.
Researchers from Tsinghua University and Georgia Tech are exploring ways to detect vulnerabilities in neural networks that use graph structured data. An Oxford University team is training multiple AI agents to efficiently operate together in the same environment. And Carnegie Mellon University researchers are determining how a neural network can more quickly learn the optimal path around a space.
Spot the Bot
All week Unsupervised.ai’s Maryam, an autonomous robot powered by NVIDIA Jetson, whizzed around the show flow delivering sweet treats.
Our social feeds were flooded with entries for our Spot the Bot competition, as attendees posted selfies with Maryam in the hope of snagging a prize. Two lucky winners went home from Stockholm with their very own NVIDIA TITAN V.
Want to discover more about the future of AI? Join us at GTC Europe from 9-11 October in Munich.