Mark Wronkiewicz

Machine learning engineer at Development Seed. Microelectronics/3D modeling/outdoors enthusiast; mathematics enthusiast in training.

[Mark Wronkiewicz @mark_wronk]

Here's a recent update on our #AI-Assist work to help people map roads networks. We experimented with deep #ReinforcementLearning to trace roads in (binarized) satellite images using a cursor 🛰️ It's was an engaging project because the algorithm traces roads like a human

[developmentseed @developmentseed]

We're exploring deep reinforcement learning (RL) to build new AI tools for tracing roads. RL has potential to be an impactful AI-assist tool within map editors since it better imitates of humans trace roads. @mark_wronk shares more on our approach.

https://t.co/TAZk5vFzWi

[Mark Wronkiewicz @mark_wronk]

After 2 years of listening, managed to slip into a @twimlai short w/ @drewbo19 to talk about our ML work at @developmentseed :D #TWIMLcon #swoon

[The TWIML AI Podcast @twimlai]

The #TWIMLcon shorts train 🚊 is back at the station! 🎙️ Check out the latest from Weiping ( @salesforce ), @drewbo19 & @mark_wronk ( @developmentseed ), @AmeenKazerouni ( @Zappos ), and @vjeffs ( @pega )
https://t.co/gjHC7DDTnu

[developmentseed @developmentseed]

Our team is growing!

We’re looking for an experienced ML engineer who’ll develop algorithms that advance our partners work in global development and address real-world social issues like the @ConnectSDGs

https://t.co/VqXrFsAiNE

[Mark Wronkiewicz @mark_wronk]

For #WorldRefugeeDay, here's an update on our #MachineLearning work to find #Refugees and #IDPs in satellite images. @Refugees just published a thorough report on scale of this problem -- 70 million people living displaced as of 2018 https://t.co/zqJJHUeIOT

[developmentseed @developmentseed]

After conflict and disasters, finding #refugees and #IDPs is a critical first step in providing humanitarian assistance. We're working with @JamonVDH's lab
@OSUCEOAS & @HOT to quickly find these vulnerable people in satellite images #WorldRefugeeDay

https://t.co/mGPrGjBeRB

[Mark Wronkiewicz @mark_wronk]

@WW2Explorer @Daks_Normandy Awesome video @WW2Explorer, glad to see everything from . another angle. We were out on the line between those two crop fields:
https://t.co/nHGIhpynFA

[Mark Wronkiewicz @mark_wronk]

.@Daks_Normandy, @IWMDuxford, forgot to tag you both. Thanks for the one-in-a-lifetime show!

[Mark Wronkiewicz @mark_wronk]

#DDay parachute drop! #DAKSoverNormandy #DDay75years #DDay75Anniversary #MissMontana
https://t.co/j3ySqQs0z0

[Mark Wronkiewicz @mark_wronk]

@tibbb @TensorFlow Thanks! Yep, we had about 7k images for training (~2k for validation) and used horizontal flipping, random brightness and contrast, and random cropping. I didn't test how performance would vary by training set size though. We'll release some of that data soon

[Mark Wronkiewicz @mark_wronk]

Here's some of our recent work to identify vulnerable housing in poor urban neighborhoods with #MachineLearning. We used object detection models (trained with @TensorFlow) to find three different building characteristics. #BuildBetterBefore

[developmentseed @developmentseed]

Homes in the world’s poor urban areas are vulnerable to natural disasters - retrofitting is cheap/effective, but finding them is hard. We’re helping the @WorldBank identify the most vulnerable homes w/ AI & street view imagery #BuildBetterBefore
https://t.co/1WL5XnNvJ1

[𝙼𝚊𝚝𝚝𝚑𝚎𝚠 𝙷𝚊𝚗𝚜𝚘𝚗 @GeoSkeptic]

. @developmentseed past and present #foss4gna San Diego @joseph_flasher @bonnie @scisco7 @mark_wronk @sharkinsgis

[Mark Wronkiewicz @mark_wronk]

Check out our new code walkthrough on deploying #MachineLearning models at scale with @Tensorflow Serving. This follows our initial work to expand electricity access (per the @UN's #SDGs) using satellite imagery

[developmentseed @developmentseed]

Deploying #MachineLearning algorithms in the real world is hard. A new tutorial covers some of what we've learned about scaling ML w/ @TensorFlow and @Docker in an effort to address the @UN's #SDGs #GlobalGoals for affordable & accessible energy for all

https://t.co/Fa5oVZlhiH

[developmentseed @developmentseed]

Come work with us! We're looking for a Cloud Engineer to join our team & help develop distributed computing models to process/analyze satellite imagery and help solve real problems with our partners like @NASA @esa @USGS that address global challenges

https://t.co/YQMldMHSVC

[Mark Wronkiewicz @mark_wronk]

Tomorrow at #AGU18 we'll be presenting our #AI powered pipeline to map the surface of #Mars that @developmentseed built w/ @tanyaofmars @heyyhannahrae from @ASU. Find us at poster 3758 #AGU18 from 8am-12pm.

Read more in our blog post:
https://t.co/L9wIh853Jy

[developmentseed @developmentseed]

Do you have plans December 13? We are hosting a happy hour w/ @Open_AQ on the heels of the AGU conference. Swing by to chat after the AGU Fall Meeting and geek out with us on Earth and space 🌏🚀🛰
https://t.co/5E4Qeeaene

[Raeed Chowdhury @raeedcho]

Check out this podcast episode I recorded with @tpbollu and @JNeurophysiol, talking about this year's #NCMSaFe meeting!

[Mark Wronkiewicz @mark_wronk]

The best in academia and industry complementing each other like salt and pepper at #CVPR2018 #CVPR18. Trying to soak in all I can

[Mark Wronkiewicz @mark_wronk]

Wrote a few of our reflections and tips for applying ML in the real world https://t.co/gTcla0dmxR

[developmentseed @developmentseed]

We documented how to set realistic goals for getting machine learning from R&D to products. @mark_wronk outlines a framework for thinking realistically about the process.

https://t.co/Crychpc9Ox

[Mark Wronkiewicz @mark_wronk]

I wonder if they tested the explosive bolts. https://t.co/ByTNhfOO0J

[Patron Saint of Pizza @dereklieu]

Spent the rainy day looking for unmapped (red) areas in #OSM, as foretold by a machine recognition algo/magic. Here's Madrid.

This is all going into Urchn, a tool to track change in urban areas. Read about it! https://t.co/IlM6CReZan @developmentseed

[Mark Wronkiewicz @mark_wronk]

RIP Tom Wolfe