Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World · HackerPapers discussions

Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World — mentioned by bravura

Wed, 21 Apr 2021 10:43:23 GMT

This is important for vision ML because sufficiently advanced simulations may comprise the real world as a special case: https://arxiv.org/abs/1703.06907 For example, gaming pixel maps have been used for semantic simulation, and have been rendered scenes: https://arxiv.org/abs/1608.02192 https://www.cv-foundation.org/openaccess/content_cvpr_2016/p... This concept has not (yet) been applied in audio ML. We have a paper in submission---will be on ArXiv soon---where we share a GPU-enabled modular synthesizer that is 16000x faster than realtime, concurrently released with a 1-billion audio sample…

Paper and discussion history. A mention is not an endorsement.

Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World — mentioned by dexter89_kp3

Thu, 04 Feb 2021 20:49:18 GMT

what is your thought on synthetic data vs active learning? For some domains, with privacy concerns or rarity of objects, getting labelled data for deep learning is challenging. There is decent research on sim2real i.e transferring models trained on synthetic data to real world applications https://arxiv.org/pdf/1703.06907.pdf

Paper and discussion history. A mention is not an endorsement.

Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World — mentioned by gdb

Sat, 01 Apr 2017 19:00:48 GMT

While the application is indeed a joke, the technique for sim-to-real transfer (and the robotic system in the post) is very real: https://arxiv.org/abs/1703.06907 :).

Paper and discussion history. A mention is not an endorsement.