DataPink

Solutions and expertise on coupling #GeoSpatial data with #Machine and #DeepLearning.

[Canwen Xu @XuCanwen]

WTF? We brutally dismember BERT and replace all his organs?
👉🏻 Check our latest work:
https://t.co/SP5gl4idsW
⛵️ BERT-of-Theseus ⛵️
*NEW MODEL COMPRESSION METHOD*
*ONE* loss + *ONE* hyperparameter + *NO* external data = GREAT PERFORMANCE
with a @huggingface -compatible weights

[DataPink @data_pink]

@data_pink will be @ParisSpaceWeek 2020

Plan to see us on our booth, for @neat_eo demo

#AI4EO

[DataPink @data_pink]

@MarcLeobet @AFBiodiversite @IGNFrance @CeremaCom @Le_Museum Humm, les aspects techniques sont déjà industrialisés.

https://t.co/3Dv3Be8fIg étant déjà utilisé en production.

Prochaine présentation demain matin @OSS_Paris

[AI at Meta @AIatMeta]

PyTorch 1.3 includes support for model deployment to mobile devices, quantization, & front-end improvements, like the ability to name tensors. New tools & libraries are also launching for improved model interpretability & multimodal development. Read more: https://t.co/HaaIt9cU3U

[David Page @dcpage3]

The paper that introduced Batch Norm https://t.co/vkT0LioKHc combines clear intuition with compelling experiments (14x speedup on ImageNet!!)

So why has 'internal covariate shift' remained controversial to this day?

Thread 👇

[Sébastien Loustau @seb_loustau]

C'est la rentrée ! Le #meetup #PAU #MachineLearning ouvre ses portes le mercredi 11 Septembre à 18h30 @Helioparc . Au menu #ComputerVision, #DeepTech et #NLP avec @o_courtin from @data_pink Inscription ICI : https://t.co/jzhdz0SOD4

[Thomas Wolf @Thom_Wolf]

I'm so excited about our current projects on energy-efficient NLP!

Distilled models are very complementary to larger models. Training cost make headlines but as large-scale models reach production, inference time will likely account for most of a model's total environmental cost

[DataPink @data_pink]

#ComputerVision ecosystem for GeoSpatial Imagery,
at scale: @RoboSatPink
https://t.co/NBeeQcCvJy

#NLP: From Text2Map, a state of art.
https://t.co/jZ4SqDEvGR

Both talks @foss4g, on how to use #DeepLearning to extract valuable patterns from #GeoSpatial (open)data.

[Jean Feydy @FeydyJean]

Optimal Transport generalizes sorting to spaces of dimension D > 1.
GeomLoss implements a smooth Quicksort-like solver, packaged as a PyTorch layer. Keeps a linear memory usage + x10 to x1,000 speedup compared with the Auction and Sinkhorn algorithms. https://t.co/HhqtuDX3TX

[Nick Weir @NickWeir09]

For the first time in the series, #SpaceNet round 5 will use a "mystery city" for final scoring! Winning algos will have to work not only on the training cities, but on a new, different geography. More details: https://t.co/BzblnOcEOh

[Planet @planet]

Next week is #ARD19, a remote sensing data interoperability workshop to be held in Menlo Park. In anticipation of the event, our senior director of analytics engineering tackles essential ingredients for ecosystem-wide geospatial analytics capability. https://t.co/xfoGSnXDg1

[Freddie Kalaitzis @alkalait]

Great news! We've open-sourced
HighRes-net for Multi-Frame Super-Resolution by Recursive Fusion.

https://t.co/9z5xvbh5UW
Our #PyTorch implementation that topped @esaACT's competition.

We're hoping this will help researchers working with satellite imagery

#AIforGood @element_ai

[DataPink @data_pink]

#TheroreticalNeuroscience in 2 links:

https://t.co/rHSAMa8Edp

https://t.co/i3LCj4nwan

[DataPink @data_pink]

@hotfigs @robhawkes @CosmiqWorks @robhawkes

Indeed, as pointed by @hotfigs, you can give a look on https://t.co/3Dv3Be8fIg tutorial:

https://t.co/CXN61TO5ZC

HTH,

[DataPink @data_pink]

@ow2 @RoboSatPink @o_courtin @Paris @OrangeGardens Related slides: https://t.co/k3R9EnJ1dy

[PyTorch @PyTorch]

A walkthrough of the PyTorch Internals by core developer @ezyang .
It's a great resource if you want to contribute to PyTorch.
https://t.co/enpD4TKjeS

[neat-EO.pink @neat_eo]

Coming https://t.co/O2w3Lf3Mmv presentations:

- May 29 #NuitCodeCitoyen @Tuba_Lyon
https://t.co/2ET1v8IZVf

- June 11 & 12 @AI_PARIS_ on @data_pink booth
https://t.co/sX3akfhWdX

- June 13 @ow2
https://t.co/aXYo6xLtVR

#ComputerVision ecosystem for #GeoSpatial #Imagery

[DataPink @data_pink]

@mouthofmorrison @developmentseed Next step could be that OpenSource Software companies already involved in GeoSpatial AI, avoid to spend energy, by keeping developing solutions by their own, and begin to converge on a common toolbox.

Could be.

cc @azavea @developmentseed @sinergise

[DataPink @data_pink]

Coming events, where @data_pink will be present:

- April 18 #AINIGHT (booth)
- May 22-23 Joint Urban Remote Sensing Event (booth)
- June 11-12 @AI_PARIS_ (booth)
- June 13 @ow2 (talk)

Hope to see you there !

[neat-EO.pink @neat_eo]

https://t.co/tdc9592C2R 0.4.0:

ChangeLog:
- Albumentation #DataAugmentation
- Improve performances
- @PyTorch JIT export
- Enhance CLI, doc and WebUI

Install: pip3 https://t.co/tdc9592C2R
Chat: https://t.co/r4RXB9agSV

#SemanticSegmentation ecosystem for #GeoSpatial #Imagery