Profile cover photo
Profile photo
Luigi Freda
65 followers
65 followers
About
Posts

Post has attachment

Post has shared content
We are excited to announce the new release of S-PTAM system corresponding to our recent paper "S-PTAM: Stereo Parallel Tracking and Mapping."

The code is available in
https://github.com/lrse/sptam

S-PTAM in action:
https://www.youtube.com/watch?v=ojBB07JvDrY

You can find the paper on http://www.sciencedirect.com/science/article/pii/S0921889015302955
Add a comment...

Post has shared content
In this new work from the Dyson Robotics Lab at Imperial, led by Stephen James and +Edward Johns, we show learned closed-loop visuomotor control for the multi-stage task of picking up an object and dropping it in a basket. At runtime, the only input to a network is the video from a single observing camera, and the output is directly joint velocities. The training is carried out entirely in simulation, with domain randomisation (random scene colours, textures and scene configuration) enabling direct transfer to the real-world demo. I find this cool and impressive, and one more of those deep learning things which I never quite expected to work until I saw it running. Most interesting to me is the closed-loop control which running a CNN at video input frame-rate allows, meaning that we can get away with a lot in terms of generalisation I think.

This paper will be presented at CoRL, the new Conference on Robot Learning to be held in Mountain View in November. Pre-print available here:
https://arxiv.org/abs/1707.02267
Add a comment...

Post has shared content
If you are hanging around Piombino tonight for the European Robotics League do not forget to visit Romeo and Delta from TRADR project!
Photo
Add a comment...

Post has attachment
https://www.wired.com/story/apples-neural-engine-infuses-the-iphone-with-ai-smarts?mbid=nl_091317_daily&CNDID=%%CUST_ID%%
Add a comment...

Post has attachment

Post has attachment
A nice trip between emergence and reductionism.

"Where in nature do you find beauty, and where do you find garbage?"

https://www.wired.com/story/how-to-solve-the-biggest-mystery-in-physics?mbid=nl_091017_daily&CNDID=%%CUST_ID%%
Add a comment...

Post has shared content
One more proof (if necessary) that computer vision is the most mature business app of Deep Learning (thanks again to ConvNets / CNNs)... That's a demo setup of a startup called Standard Cognition, which is using a network of cameras and DL powered machine vision to create a store without cashiers.
Add a comment...

Post has attachment
"NVIDIA’s autonomous mobile robotics team today released a framework to enable developers to create autonomous drones that can navigate complex, unmapped places without GPS. All of this is done through deep learning and computer vision powered by NVIDIA Jetson TX1/TX2 embedded AI supercomputers."

https://news.developer.nvidia.com/nvidia-researchers-release-trailblazing-deep-learning-based-framework-for-autonomous-drone-navigation/
Add a comment...

Post has shared content
Wait while more posts are being loaded