Maxim Neumann

Explore the incredible work from @GoogleResearch and @GoogleDeepMind who are leveraging AI to help understand and protect our natural spaces:

@GoogleResearch @GoogleDeepMind Thrilled to see this live! Huge thanks to the team.

Together with @GoogleDeepMind, we present Natural Forests of the World 2020, a new map & dataset that distinguishes natural forests from other tree cover with 92.2% accuracy, creating a 10m baseline to support global conservation efforts. Learn more: goo.gle/49ePZcF @WorldResources @IIASAVienna

Our PaliGemma technical report is finally out: arxiv.org/abs/2407.07726. We share many insights that we learned while cooking the PaliGemma-3B model. Both about pretraining and transfer.

PaliGemma also includes two remote sensing VQA tasks: RSVQA-HR and RSVQA-LR (by @SylvainLobry, @devistuia). Without anything RS specific, it directly achieved SOTA, improving upon RSGPT or SHRNet.

@sarameghanbeery @aihelpsukraine Since both topics are mentioned at once (Ukraine help and urban trees), have a look at recently released urban tree maps over Kiev and Odessa (seems to be only in Ukrainian): - successful-city.com/zapustyly-onla… - #map=13//46.47546295270333//30.730133056640625&&layer=1664656129424098568-1,100//2521278728567260649-1,100//2521267671635330532-1,100//2263127208933459762-1,100&&action=info-2263127208933459762,2412678150216484399" target="_blank" rel="nofollow noopener">greencity.omr.gov.ua/dereva?fbclid=… Maybe it's a direction for AutoArborist?

Starting soon! @aihelpsukraine is a charity conference on AI to raise funds for medical and humanitarian aid for Ukraine. Donate if you can, and tune in to hear about mapping urban forests with AI, multiview computer vision, and the challenges of geospatial distribution shifts

Stop by the Google booth at #ECCV2022 at 3:30 pm today to see a demo presented by Austin Stone, @MJLM3 and @agritsenko about OWL-ViT, a simple and scalable approach for open-vocabulary object detection and image-conditioned detection. Try it yourself at bit.ly/owl-vit-demo.

Today at 5.30pm (CET) / 8.30am (PST), @rikelhood @_basilM and myself will be presenting our poster on "Scaling Vision with Sparse Mixture of Experts" at the @NeurIPSConf. neurips.cc/Conferences/20…

Today we are releasing the Open Buildings Dataset, a new open-source dataset containing the locations and footprints of >500M buildings with coverage across Africa, which can support numerous scientific and humanitarian applications. Read more at goo.gle/3751HVR

[3/3] Towards big vision While dense models are still the norm, sparse MoE layers can work well too! Large Vision-MoEs (15B params) can be trained to high performance relatively efficiently, and can even prioritize amongst patches (see duck). arxiv.org/abs/2106.05974 ...

Ever-larger computer vision models - by now with 15 billion parameters using sparse mixture-of-experts for reduced computation and achieving 90.35% on ImageNet!

This book was long time in making... I wrote on a chapter of it 8 years ago :)

Happy to announce that our paper “Scalable Transfer Learning with Expert Models” has been accepted to ICLR 2021! Joint work with @rikelhood @_basilM @rengglic @ASusanoPinto @sylvain_gelly @keysers @neilhoulsby openreview.net/forum?id=23ZjU…

The wildfire situation in California and Oregon has now escalated to the point that I can no longer keep track of the countless massive, fast-moving, and potentially very dangerous fires. The geographic scale and intensity of what is transpiring is truly jarring. #CAwx #ORwx

Really cool project - great to see this initiative moving forward and getting even Al Gore on board! sprou.tt/1CFtfVVOjiW

arxiv.org/abs/2006.16668 We scaled the Transformer model with Sparsely-Gated Mixture-of-Experts using GShard, and trained a 600B multilingual translation model in about 4 days (for 100 languages) achieving 13.5 BLEU gain compared to the baseline.

#Spacenet6 Challenge: Extracting building footprints using Multi-Sensor All Weather Mapping 🆕#Opensource #remotesensing datasets of half-meter #SAR @capellaspace and half-meter electro-optical imagery @Maxar #computervision #DataScience ➡️More at cutt.ly/0rcyFBf

We knew #ClimateChange would cause mass migration of people. Not sure we predicted that scores of terrified #kangaroos would seek safe haven in residential streets. This is reality now. But still, let’s protect the coal industry, eh. #AustraliaBurning twitter.com/backt0nature/s…

👇(Thread) Some Advice for Ph.D. Students:👇 💫ideas: the best and most interesting research ideas came to my mind when I was either cycling or taking a shower (seriously!); 💫ideas: read old papers as well, find out what scientists debated; #PhDChat #phdlife

10 ML & NLP Research Highlights of 2019 New blog post on ten ML and NLP research directions that I found exciting and impactful in 2019. ruder.io/research-highl…