Antonio Loquercio @antoniloq Twitter profile

Pinned Tweet

Antonio Loquercio

2 years

We train a robot 🐕 to traverse complex terrains with a monocular RGB camera from its own real-world experience! To do so we propose Cross-Modal Supervision (CMS), an algorithm to supervise vision using proprioception. Project Page: 1/5

21

56

329

Last Seen Profiles

@ela007a

@Soontara584130

@t3rrified

@cat_luru

@PandezLabs

@KatelinP899

@merr_marke

@nakayamasimasi

@PilarGarrido_

@stw_pdg

@_YamSmalls_

@tamukaz

@Xa1JeSs7bGCyRB9

@TEKKAMAKI3

@jandakembangstw

@Kokeshichan11

@Milazzo__7

@Lucydragonex

@GiovanniDElia11

@servelunablack

@NikilisRBX

@sleazy_art

@KathlynnDeHaan

@MaidenViking_

@ShadowReindeer

@DannyTheHope

@stefanziller

@Apexerkarma

@t_miskovi

@shan_gram95

@idrissa_no81449

@suiseichuchu

@saji2423

@KrunchyCroc

@MonaQueen__

@beaushines

Antonio Loquercio

@antoniloq

9 months

Excited to share that I'll join UPenn as an Assistant Professor next fall! I couldn't be more grateful to all the mentors and collaborators for their support along the way! I'm looking for motivated students👩‍🎓to join my lab! If you love robots 🤖 and learning, look no further!

29

27

423

Antonio Loquercio

@antoniloq

2 years

Honored to receive the Georges Giralt PhD Award 🥇 (awarded to Europe's best thesis in Robotics 🤖) for my work on #learning vision-based high-speed #drone flight ✈️! Thanks to @davsca1 for the great supervision and all my collaborators! Thesis: @UZH_en

9

11

283

Antonio Loquercio

@antoniloq

1 year

Happy to share that my Ph.D. thesis on Agile Vision-Based Flight has been published by @Springer Tracts in Advanced Robotics (STAR)! 🌟 I've tried to summarize all the lessons I've learned over 5 years to unlock agile and vision-based flight. 📷 🚁✨

4

12

212

Antonio Loquercio

@antoniloq

3 years

📢 Personal Update 📢: I'm excited to join UC Berkeley for a PostDoc with Prof. Jitendra Malik. I'll continue to work at the intersection of computer vision and robotics, and I am proud to have the opportunity to contribute to the next generation of intelligent robots 🤖

5

2

141

Antonio Loquercio

@antoniloq

1 year

Just released the code for my legged locomotion project: Excited to see what others can do with it! #robotics #opensource

GitHub - antonilo/vision_locomotion: Project Code for the paper "Learning Visual Locomotion with...

Project Code for the paper "Learning Visual Locomotion with Cross-Modal Supervision" (ICRA2023) - antonilo/vision_locomotion

github.com

Antonio Loquercio

@antoniloq

2 years

We train a robot 🐕 to traverse complex terrains with a monocular RGB camera from its own real-world experience! To do so we propose Cross-Modal Supervision (CMS), an algorithm to supervise vision using proprioception. Project Page: 1/5

21

56

329

2

20

124

Antonio Loquercio

@antoniloq

10 months

Only five years ago, @kaufmann_elia and I were proud of having vision-based drone racing at walking speed. After years of hard work (by us and many other people in the community), we achieved super-human performance. Next challenge: generalization Paper:

3

16

123

Antonio Loquercio

@antoniloq

2 years

They told me to follow in my advisor's footsteps. To be more efficient, I've coded my robot to do it for me 🤖 Starring prof Jitendra Malik. Ph. Credits: @akanazawa

1

3

95

Antonio Loquercio

@antoniloq

4 years

Very motivated by the Plenary Talk of @ylecun at #ICRA2020 ! What he identifies as the "next big challenges" of #Robotics is exactly the core of our research at the Robotics and Perception Group in @UZH_en !

3

16

63

Antonio Loquercio

@antoniloq

3 years

Check out our #ICRA2021 workshop on perception and action in dynamic environments! We have amazing speakers and organized a competition for autonomous drone navigation in a forest (Prize is a Skydio2)! w/ @davsca1 , @lucacarlone1 , @MarkusRyll

0

9

58

Antonio Loquercio

@antoniloq

1 year

It was great to spend the afternoon in #Stanford at @drmapavone ’s lab and give a talk about my recent work on #robot 🤖 learning. Got plenty of new ideas speaking with many interesting and smart people there, looking forward to coming back soon!

3

1

52

Antonio Loquercio

@antoniloq

4 years

#Flightmare accepted to the conference on Robot Learning #CORL2020 🎉🍾 Integrating reviewers’ comments and soon releasing an improved version 💪 #UZH #ETH #drones #Simulation

Davide Scaramuzza

@davsca1

4 years

Are you looking for a photo-realistic, customizable, and easy to use simulator for #drones ? Then, #Flightmare is what you need! Compatible with #ROS , #OpenAI Gym, and #Oculus . Website: Kudos to @Cprimer_ @kaufmann_elia @antoniloq #UZH_en

1

51

178

0

4

49

Antonio Loquercio

@antoniloq

3 years

Check out our #ICRA2021 workshop on Perception and Action in Dynamic Environments: We have an amazing line of speakers, take the opportunity to join to ask questions! Workshop will be on Friday, June the 4th! w/ @davsca1 @lucacarlone1 @MarkusRyll

0

8

38

Antonio Loquercio

@antoniloq

1 year

What is the role of data in robotics? What would a robotics foundation model look like? 🤖 If you're interested in these questions, join us tomorrow for the #ICRA23 "Pre-training for Robotics" workshop! 🚀 More information on the workshop website:

0

2

35

Antonio Loquercio

@antoniloq

1 year

Honored to receive the Frontiers of Sciences Award for my work "Learning High-Speed Flight in the Wild" (Science Robotics, 2021) w/ @kaufmann_elia @ranftlr1 M. Mueller V. Koltun and @davsca1 !

2

0

30

Antonio Loquercio

@antoniloq

1 year

Autonomous drone racing is becoming a larger and larger field. In this work, we organize existing works and predict the next challenges for the community. We hope that you find it useful!

Davide Scaramuzza

@davsca1

1 year

Autonomous systems are revolutionizing industries from surveying to delivery. But to navigate as agile as humans, we need robust algorithms. Learn about the progression of this field and the challenges ahead in our new survey on #autonomousdroneracing

3

37

144

0

2

28

Antonio Loquercio

@antoniloq

3 years

Thanks to my advisor for guiding me in this wonderful journey and to the reviewers for asking insightful questions! #DeepLearning for #drone navigation is an amazing topic to work on!

Davide Scaramuzza

@davsca1

3 years

Congratulations to Antonio Loquercio @antoniloq for successfully defending his PhD on “Agile Autonomy: Learning High-Speed Flight”! Thanks to the reviewers @rsiegwart @angelaschoellig @pabbeel ! Video of defense: #DeepLearning #robotics #sim2real

1

3

78

1

0

29

Antonio Loquercio

@antoniloq

5 years

Check out our T-RO paper on simulation to real world transfer for drone racing: Trained in a non-photorealistic simulation and deployed on a real quadrotor without any fine-tuning. w/ @davsca1 #UZH #UZH_en #droneracing #sim2real #Robotics #DeepLearning

0

4

23

Antonio Loquercio

@antoniloq

2 years

In our latest work, we pushed the limits of adaptive control for quadcopters. It turns out that a single policy (trained in simulation) can fly multiple drones while rejecting disturbances! w/ @ashishkr9311 , @mark_w_mueller and lab! Project Page:

2

0

25

Antonio Loquercio

@antoniloq

3 years

Truly humbled by this nomination! In this work, we learnt that sim2real really works for agile navigation! So much has happened since then ✈️

Davide Scaramuzza

@davsca1

3 years

Honored that our IEEE Transactions on Robotics 2020 paper "Deep Drone Racing: From Simulation to Reality with Domain Randomization" was selected Best Paper Award finalist! w/ @antoniloq @kaufmann_elia PDF, video, code:

1

3

89

1

0

24

Antonio Loquercio

@antoniloq

4 years

One of coolest work I've ever been part of. Can't really describe the feeling we got when our control policy, trained only in simulation, was flying a real quadrotor with such agility!

Davide Scaramuzza

@davsca1

4 years

Drones with on-board sensing and computation can now fly agile acrobatic maneuvers! Check out our #RSS2020 paper "Deep Drone Acrobatics"! Paper/Code: Video: Kudos: @antoniloq @kaufmann_elia @UZH_en #Intel #MachineLearning #drones

4

49

150

0

1

20

Antonio Loquercio

@antoniloq

10 months

While previous work mainly studied multi-agent interaction in controlled environments and state observations, @andrea_bajcsy and I were interested in seeing what happens when deploying "in the wild" with only camera observations. We find that constraints bring about creativity!

Andrea Bajcsy

@andrea_bajcsy

10 months

Ever wonder why robots “in the wild” are always alone? Our learning approach enables robots to information gather, predict, and strategize about other agents in unstructured environments, all from an onboard camera. Project page:

10

31

205

1

2

22

Antonio Loquercio

@antoniloq

2 years

🤯 first time I see a robot walking on stepping stones in the real world. This is just amazing! 🔥

Deepak Pathak

@pathak2206

2 years

After 3yrs of locomotion research, we report a major update in our #CoRL2022 (Oral) paper: vision-based locomotion. Our small, safe, low-cost robot can walk almost any terrain: high stairs, stepping stones, gaps, rocks. Stair for this robot is like climbing walls for humans.

34

160

895

0

1

21

Antonio Loquercio

@antoniloq

4 years

Tomorrow 4PM CET (10AM ET) I will talk about my work on Agile Flight at the ETH Autonomy Talks Series. Join if you would like to know the secrets of controller learning! Registration is totally free! --> #Robotics #drones #MachineLearning

1

4

20

Antonio Loquercio

@antoniloq

4 years

Very excited to see @derspiegel covering our work on #drones #acrobatics with #AI in its latest printed edition! Check out the video if you haven't already seen it!

1

3

18

Antonio Loquercio

@antoniloq

2 years

Enjoyed reading this piece. An antidote to the current "foundation model can do it all" philosophy. I get proof of this every day: I have no idea how to tell my robot, in any (programming) language, how to climb up the stairs without falling 🤖

Yann LeCun

@ylecun

2 years

New piece by @Jake_Browning00 & me: - language carries a small portion of all human knowledge. - much of human knowledge & all of animal knowledge is non verbal (& non-symbolic). - hence large language models cannot come close to human-level intelligence.

53

265

979

1

5

15

Antonio Loquercio

@antoniloq

2 years

Register for our #iros2022 Safe Robot Learning competition! Develop safe learning-based algorithms for quadrotors that can cope with uncertainties not known at design time. Website: Codebase: Video:

0

14

Antonio Loquercio

@antoniloq

3 years

We had so much fun doing this project! Agile flight is challenging, but lot of fun!

Davide Scaramuzza

@davsca1

3 years

An end-to-end policy trained in simulation flies vision-based #drones *in the wild* at up to 40 kph! Check out our new @SciRobotics paper. Code released! w/ @antoniloq @kaufmann_elia Video: Paper, Code & Datasets: @UZH_en #Intel

6

40

189

0

1

13

Antonio Loquercio

@antoniloq

3 years

Our work on high-speed drone navigation is on the landing page of @ScienceMagazine (). Such an honor to be there together with amazing works from all disciplines of science! 😁

1

12

Antonio Loquercio

@antoniloq

4 years

Our deep #drone #acrobatics has been nominated for the Best Paper Award in #RSS2020 ! Very honored and flattered at the same time for this nomination!

Davide Scaramuzza

@davsca1

4 years

Very honored for being nominated for the #RSS2020 Best Paper and Best System Paper Awards! Congratulations to my students and collaborators and all the other finalists! @uzh_ifi @nccrrobotics @UZH_en

1

4

75

0

1

10

Antonio Loquercio

@antoniloq

2 years

Our DodgeDrone competition was a big success! We had a large number of submissions and an exciting final on stage at #ICRA2022 ! Congratulations to the winners! We even had real drones flying around 🤖

0

10

Antonio Loquercio

@antoniloq

2 years

Drones are saving people's lives, *today*! That's why research on (agile) aerial robotics is so important! Scaling these systems up will require drones to be autonomous.

Reuters

@Reuters

2 years

A pioneering drone lifeguard service rolled out across Spanish beaches saved the life of a 14-year-old boy this month as he struggled against a powerful tide off a beach in Valencia, its operators said

6

92

318

0

10

Antonio Loquercio

@antoniloq

2 years

The better the vision, the better the walking. After just 30min of real-world data, our policy can traverse complex terrains like stairs (up to 19cm high), curbs (up to 20cm high), slopes, dirt roads, and construction sites. W/ @ashishkr9311 and @JitendraMalikCV 5/5

0

9

Antonio Loquercio

@antoniloq

4 years

"Quadrotors are among the most agile and dynamic machines ever created." Thanks to @BotJunkie for the great article on our #RSS2020 work on drone acrobatics! It was lot of fun to be interviewed 😎

AI-Powered Drone Learns Extreme Acrobatics

This drone can perform maneuvers that are challenging for even the best human pilot

spectrum.ieee.org

0

3

10

Antonio Loquercio

@antoniloq

3 years

Impressive memory savings for very high-frame-rate videos: from 16GB to 0.4GB. We can eventually film #agile #drones for more than a few seconds :)

Davide Scaramuzza

@davsca1

3 years

We release our newest work on event cameras: "Time Lens". We use events to upsample low-framerate RGB HD video by over 50 times with only 1/40th of the memory footprint! #CVPR2021 Paper, code, datasets: @DanielGehrig6 @MathiasGehrig

12

196

897

0

9

Antonio Loquercio

@antoniloq

2 years

The terrain geometry below the robot is observable from a history of leg positions and orientations (i.e. proprioception). Time-shifting proprioception estimates from previous real-world experiences, we train a visual module to predict the terrain geometry ahead of the robot. 2/5

1

0

8

Antonio Loquercio

@antoniloq

3 years

The very confident @kaufmann_elia and I did another amazing paper together. What can it be about? 🤔 Get ready for a mind-blowing video 🤯🤯🤯

0

8

Antonio Loquercio

@antoniloq

1 year

I use it on my robot daily. Strongly recommended! Thanks for sharing @philippswu

Philipp Wu

@philippswu

1 year

In Daydreamer, we found it helpful to protect the robot in situations like this👇. Sharing the turtle shell I designed and hope it will be useful to other researchers!

5

11

151

1

0

9

Antonio Loquercio

@antoniloq

4 years

Are you interested in geometric deep learning and 3D mesh processing? Then check out our latest #NeurIPS2020 paper!

Luca Carlone

@lucacarlone1

4 years

2/2) "Primal-Dual Mesh Convolutional Neural Networks" (+ Francesco Milano, Davide Scaramuzza, Antonio Loquercio, Toni Rosiñol Vidal) - Paper: - Video: (coming soon!) - Code: #mitsparklab #robotPerception #learning

0

8

0

1

8

Antonio Loquercio

@antoniloq

4 years

Took it way more than I expected, but I am eventually done with #neurips2020 reviews! The recent tutorial on "How to Write a Good Review" () significantly improved my review quality. Thanks again to @lealtaixe and @SattlerTorsten for organizing it!

0

8

Antonio Loquercio

@antoniloq

3 years

A very nice article about our work from Forbes. Nice how journalists can explain complex concepts in simple words :)

Davide Scaramuzza

@davsca1

3 years

Very honored to see our @SciRobotics paper featured in Forbes! Paper, code, video: @UZH_Science @UZH_en @antoniloq @kaufmann_elia @UZHspacehub

0

6

42

0

8

Antonio Loquercio

@antoniloq

5 years

Check out our paper on Uncertainty Estimation in #DeepLearning ! The framework we propose is flexible and offers provable guarantees on estimation!

Davide Scaramuzza

@davsca1

5 years

Check our new paper on a General Framework for Uncertainty Estimation in Deep Learning. PDF: Video: e/ @antoniloq #UZH #UXH_en #DeepLearning #Robotics

0

7

34

0

1

8

Antonio Loquercio

@antoniloq

2 years

Looking forward to meeting colleagues and friends in Phil. next week! I will give three keynote talks and present my paper on agile flight! Please reach out if you want to chat!

Davide Scaramuzza

@davsca1

2 years

We are looking forward to presenting these 9 papers on perception, learning, planning, and control in person next week at @ieee_ras_icra ! Additionally, we will be presenting in many workshops. Full list with links, times, and rooms:

0

8

29

0

8

Antonio Loquercio

@antoniloq

2 years

Next week I'll talk about the intricacies of #vision + #robotics in the real world at #Ellis Turin. What can we roboticists learn from decades of human vision theories? Join if you want to know! Cool videos guaranteed :)

Tatiana

@tommasi_tatiana

2 years

Let's inaugurate the Ellis Turin Talk series! February 9th, Antonio Loquercio (UC-Berkeley) Check all the details and register at

0

1

8

0

1

8

Antonio Loquercio

@antoniloq

1 year

The results are mind-blowing! I think this opens a whole new set of possibilities for studying human behavior and learning from them!

Jitendra MALIK

@JitendraMalikCV

1 year

We can now reconstruct 3D humans from single images (forthcoming, ICCV 2023). Model available to try on your images / videos at

6

45

360

0

8

Antonio Loquercio

@antoniloq

2 years

We spent quite a lot of time ⌛️building an API that is powerful and easy to use. You need ZERO prior knowledge of drones ✈️ to do well in this competition! Check it out!

Davide Scaramuzza

@davsca1

2 years

Do you want to push the boundary of aerial robotics? Participate in our #icra2022 #DodgeDrone Challenge! . We provide an easy-to-use API and a #ReinforcementLearning framework! The winner will get a talk invitation and a money prize!

1

23

111

1

0

8

Antonio Loquercio

@antoniloq

2 years

CMS naturally enables continual learning from experience. Every day, we use the latest vision policy to walk, while proprioception constantly estimates the geometry in the background. This in turn generates more data, on which we train a new policy overnight. 4/5

1

0

6

Antonio Loquercio

@antoniloq

4 years

Our project was covered in the @MailOnline Tech news! Check it out!

Navigation algorithm lets drones perform challenging acrobatics

Scientists have developed a quadrotor helicopter, or quadcopter, that can learn to fly acrobatic manoeuvres that would challenge even a human operator.

www.dailymail.co.uk

0

6

Antonio Loquercio

@antoniloq

2 years

Amazing to see people making progress on our competition!

Mandela Ouafo

@mandel98

2 years

#DodgeDrone challenge icra 2022 First algorithm design. @antoniloq

2

0

11

1

0

7

Antonio Loquercio

@antoniloq

1 year

@JitendraMalikCV gave a great talk today, discussing why robotics is a necessary milestone towards AI. Punch line: "intelligence is much older than us humans". Honored to have my work featured during his talk! Here is a recording if you missed it:

0

2

7

Antonio Loquercio

@antoniloq

11 months

@ashishkr9311 is one of the people I learned most during my stay at Berkeley. The @Teslasbot has a much higher chance of success with him on the team!

Ashish

@ashishkr9311

11 months

Just graduated from @UCBerkeley , now at @Tesla ! In my PhD, I showed the effectiveness of pure learning for: -Locomotion () -Dexterous manipulation () -Universal drone flight () Next up: humanoids! ( @Teslasbot )

37

39

833

0

6

Antonio Loquercio

@antoniloq

4 years

@gabepsilon @Jan_R_Peters @DeepMind I totally understand this line of research. However, calling this "real world" is a bit disrespectful to all people working with the *real* world. As you're sending the wrong message to the community, please consider re-naming!

1

0

6

Antonio Loquercio

@antoniloq

4 years

Check out this work if you want to know how to decrease the computational complexity of your event camera network up to 20 times! TL;DR: Use the properties of your data to improve efficiency.

Davide Scaramuzza

@davsca1

4 years

Bring back the sparsity in event-data using our Event-based Asynchronous Sparse Convolutional Networks! Kudos to @Messikommer1 , @DanielGehrig6 , @antoniloq Paper: Code: Video: #ECCV2020 #eventcameras

1

2

18

0

1

7

Antonio Loquercio

@antoniloq

3 years

Join us for the final industrial session of our workshop () with @wolfram_burgard @RaquelUrtasun and Richard Newcombe!

0

6

Antonio Loquercio

@antoniloq

2 years

I can't recommend more working with @guillermogb . I have really learnt a lot from him. Join his lab if you're serious about Computer Vision!

Guillermo Gallego

@guillermogb

2 years

8 PhD positions at HEIBRiDS #DataScience Graduate School in Berlin. Apply now! @ECDigitalFuture @helmholtz_de

0

4

9

0

6

Antonio Loquercio

@antoniloq

4 years

Do you want to know how to make autonomous drones fly crazy maneuvers? Check out the blog post below 👇🏾

sim2real

@sim2realAIorg

4 years

In this post, @antoniloq and @kaufmann_elia explain how they use simulations for their work on deep drone acrobatics.

0

15

59

0

3

6

Antonio Loquercio

@antoniloq

4 years

Code coming soon, stay tuned 🤖

Davide Scaramuzza

@davsca1

4 years

Can we learn a 3D perception system with the same inputs as an human, i.e. vision and sparse depth measurements? Yes! To know how, see our paper by @antoniloq & @Google "Learning Depth with Very Sparse Supervision" PDF: Video:

1

16

113

0

1

6

Antonio Loquercio

@antoniloq

1 year

Strongly recommended for everybody who is serious about Computer Vision. I met @dimadamen at Berkeley, and can confirm she's not only an outstanding researcher but also an amazing advisor!

Dima Damen

@dimadamen

1 year

[Pls RT 📢 - DL 23/04] Fantastic 💡 **postdoc** Opportunity to join my group @BristolUni working around multimodal video understanding. Project in collab w Andrew Zisserman &others @Oxford_VGG . 26 months w student co-supervision opportunities. Email for queries or apply directly.

1

34

81

0

6

Antonio Loquercio

@antoniloq

2 years

However, this requires a walking policy in the first place. So how to break this chicken-and-egg problem? A robot can walk (clumsily of course) when blind, so the bootstrapping can start from there. 3/5

1

0

5

Antonio Loquercio

@antoniloq

2 years

Academia version: Ph.D. student after writing the first three words of their PI's grant proposal.

🌒 Eclipse Java IDE

@EclipseJavaIDE

5 years

Developer after writing the first three words documenting their software.

3

74

205

0

5

Antonio Loquercio

@antoniloq

2 years

@gautamcgoel But it's also an insanely beautiful job :)

1

0

4

Antonio Loquercio

@antoniloq

5 years

Our work on Nano-Drone Navigation has just been published on IEEE Xplore! Really excited about future work in this direction! Video: Paper:

0

5

Antonio Loquercio

@antoniloq

3 years

Sim2Real requires accurate dynamic models. That's probably the best you get for #drones !

Davide Scaramuzza

@davsca1

3 years

Check out our #RSS2021 paper "NeuroBEM: Hybrid Aerodynamic Quadrotor Model". We augment a first-principle approach with #deeplearning to achieve unprecedented modeling of fast quadrotor flight. Paper, Code, Dataset: @l_bauersfeld @kaufmann_elia @foehnph

1

18

72

1

0

5

Antonio Loquercio

@antoniloq

3 years

@RosinolToni 2025: Slam no more needed, AGI has a spatial understanding that is no more metric 😂 Of course joking 🙃

2

0

5

Antonio Loquercio

@antoniloq

1 year

Bottom-up feels the right way! But time will tell...

Autonomous Vision Group

@AutoVisionGroup

1 year

Vladlen delivering the opening statement for the 'against' team.

2

40

0

3

Antonio Loquercio

@antoniloq

3 years

Terrific work by my colleague @realyunlong !

Davide Scaramuzza

@davsca1

3 years

Autonomous Drone Racing with Deep Reinforcement Learning! Check out the paper and the amazing video, and turn up the volume ;-) Real-world experiments at the end! PDF: Kudos to @realyunlong @kaufmann_elia

0

6

31

0

4

Antonio Loquercio

@antoniloq

4 years

Office view for today!

0

4

Antonio Loquercio

@antoniloq

3 years

Impressive work on how to control drones when one propeller fails! This work is the result of an amazing collaboration between @UZH_en and @tudelft !

NCCR Robotics

@nccrrobotics

3 years

How to keep drones flying when a motor fails? Researchers at the @UZH_en and @TUDelftDRI show how onboard cameras can be used to keep damaged quadcopters in the air and flying stably – even without GPS. @davsca1

0

5

10

0

2

4

Antonio Loquercio

@antoniloq

2 years

A big congrats also to all other finalists and the ex-equo winner @_mlutter ! All theses were really outstanding! Awesome to see so much great robotics research 🤖 happening in Europe!

1

0

4

Antonio Loquercio

@antoniloq

2 years

We now operate at relatively low speed, in close to hovering conditions. Stay tuned for higher speeds ✈️

1

0

4

Antonio Loquercio

@antoniloq

2 years

Quite interesting findings in here! Noise helps in low-data regimes much more than I expected. We should try this for vision/robotics tasks.

Noëmi Aepli

@noeminaepli

2 years

A simple yet effective strategy to improve cross-lingual transfer between closely related language varieties 👀 an #acl2022nlp findings paper with @RicoSennrich . Pre-print:

7

28

164

0

4

Antonio Loquercio

@antoniloq

4 years

If you work on #drones , #ReinforcementLearning , or #ComputerVision , you should try this out 👇

Davide Scaramuzza

@davsca1

4 years

Are you looking for a photo-realistic, customizable, and easy to use simulator for #drones ? Then, #Flightmare is what you need! Compatible with #ROS , #OpenAI Gym, and #Oculus . Website: Kudos to @Cprimer_ @kaufmann_elia @antoniloq #UZH_en

1

51

178

0

1

4

Antonio Loquercio

@antoniloq

1 year

@kevin_zakka Same feeling every single day :/ (Love your project btw, it’s insane it has no citation). But at some point people will realize that you must do robotics if you want to do robotics. Even thought at times language can be extremely helpful:

0

4

Antonio Loquercio

@antoniloq

4 years

Answering some questions on Agile Flight at the Italian conference of Robotics and Intelligent Machine (I-RIM3D). Feel free to join if you're curious about our work:

0

1

4

Antonio Loquercio

@antoniloq

5 years

Very proud of this work!

Kosta Derpanis

@CSProfKGD

5 years

Our new work "End-to-End Learning of Representations for Asynchronous Event-Based Data", Daniel Gehrig @DanielGehrig6 , Antonio Loquercio @antoniloq , Konstantinos G. Derpanis, Davide Scaramuzza @davsca1 arXiv: video:

1

3

18

0

2

Antonio Loquercio

@antoniloq

4 years

@CSProfKGD Accepted!!!! 🎉🎈💥 Details and preprint coming soon!

1

0

4

Antonio Loquercio

@antoniloq

4 years

We are happy to answer all your questions on our #NeurIPS2020 paper on 3D Mesh processing in gather town! Please reach out! 👇👇👇

Antoni Rosinol

@RosinolToni

4 years

Join us at NeurIPS to learn how to learn on 3D Meshes, manifold or not! (+ Francesco Milano, @antoniloq , @davsca1 , @lucacarlone1 ) Link in comments! - Paper: - Code: - Video:

1

13

0

4

Antonio Loquercio

@antoniloq

4 years

Looking forward to this workshop today! If you are interested in perception, learning, or control you should join it, it's totally free!

Davide Scaramuzza

@davsca1

4 years

Join today our final day of the #IROS2020 workshop "Perception, Learning, and Control for Autonomous Agile Vehicles" organized by myself, @loiannog and @SertacKaraman . From 10 am to 2 pm EST time. Homepage: Today’s program below:

0

3

18

0

1

4

Antonio Loquercio

@antoniloq

1 year

The backbone of our research on agile robotics is now open access. Agile flight is democratized!

Davide Scaramuzza

@davsca1

1 year

Ready to take your autonomous quadrotor research to the next level? We are excited to release #Agilicious , our open-source and open-hardware platform for agile flight! It supports both model-based and neural-network controllers! @uzh_en @SciRobotics

2

15

51

0

4

Antonio Loquercio

@antoniloq

2 years

Submission is now open! Check the details here: . We will be announcing the finalist in 10 days! w/ @davsca1 @realyunlong @kaufmann_elia

GitHub - uzh-rpg/agile_flight: Developing and Comparing Vision-based Algorithms for Vision-based...

Developing and Comparing Vision-based Algorithms for Vision-based Agile Flight - uzh-rpg/agile_flight

github.com

Davide Scaramuzza

@davsca1

2 years

Do you want to push the boundary of aerial robotics? Participate in our #icra2022 #DodgeDrone Challenge! . We provide an easy-to-use API and a #ReinforcementLearning framework! The winner will get a talk invitation and a money prize!

1

23

111

0

4

Antonio Loquercio

@antoniloq

3 years

Dissemination is a fundamental part of research! Some very good advice in here 👇

Jia-Bin Huang

@jbhuang0604

3 years

Sharing ideas on how to disseminate your research. "I am THRILLED to share that our paper is accepted to ..." Congrats! So what's next? No one is going to browse through the list of thousands of accepted papers. Ain't nobody got time for that. Check out 🧵below for examples.

11

181

729

0

1

4

Antonio Loquercio

@antoniloq

1 year

Quite fun chat with @lady_nerd :) Enjoyed the conversation, and additionally got a new perspective on robotics research 🤖

Build Amazing Things (securely) Podcast

@build_amazing

1 year

🎙️ New episode! 🎙️ Antonio Loquercio is a roboticist and academic from UC, Berkeley. In this episode, Laura and Antonio dig into the current state of autonomous robots, the challenges of making them navigate environments safely, and where this exciting

0

4

Antonio Loquercio

@antoniloq

2 years

@zhendongsu @chengyuzh @RiggerManuel @theodortheod @ast_eth @ASPLOSConf @CSatETH congrats!

0

4

Antonio Loquercio

@antoniloq

2 years

@AlexGDimakis I am not sure I agree... All systems you mentioned have some serious failure cases. Dalle2 for example can't get spatial relationships or attribution predicates. With the current techniques we have, solving these problems requires clean data (i.e. human work), not large data.

0

3

Antonio Loquercio

@antoniloq

5 years

Very excited to give a talk in few weeks about our reinforcement learning research on autonomous drone flight! #robotics #robots #drones #reinforcementlearning All details here:

0

3

Antonio Loquercio

@antoniloq

2 years

@lonebots @BostonDynamics This robot is approximately 5 times cheaper, much smaller, and lighter. It is perfect from a research perspective since you can change low-level code.

2

0

3

Antonio Loquercio

@antoniloq

4 years

@CSProfKGD @eccvconf @openreviewnet Totally agree with this line of thought, making the review process open has definitely served the #ICLR community well. What I don't like about it is that people stop thinking with their head, just read the reviews/AC comment and assume they got the paper.

0

1

3

Antonio Loquercio

@antoniloq

2 years

@SkydioHQ It would be amazing if you could add credits and/or sources of the images you used in this post :)

0

3

Antonio Loquercio

@antoniloq

4 years

@iamsashasax @LerrelPinto Nice work! You might be interested in our RSS2020 work deep drone acrobatics (), where we formally and experimentally show that mid-level representations (which we call abstractions) reduce the sim2real gap!

1

0

3

Antonio Loquercio

@antoniloq

10 months

@AjdDavison @xiao_ted Every part of our system builds upon decades of research (starting from the on-board computer!). We made our best to explain our choices in detail (see section The SWIFT System) and justify design choices.

0

2

Antonio Loquercio

@antoniloq

3 years

@BotJunkie @davsca1 @chelseabfinn @wolfram_burgard @SkydioHQ @RaquelUrtasun @AleksandraFaust yes, we will record all talks and make them available on our website afterward the workshop!

0

1

3

Antonio Loquercio

@antoniloq

4 years

Our work on uncertainty estimation really rocks! Code is available, try it out!

Davide Scaramuzza

@davsca1

4 years

Check out our RA-L & ICRA paper on uncertainty estimation of deep neural networks! Paper: Code: Video: w/ @antoniloq @MattiaSegu #UZH #UZH_en #DeepLearning #Robotics

0

18

79

0

3

Antonio Loquercio

@antoniloq

5 years

I had the pleasure to work with Kosta and I found him to be a great supervisor! Anybody interested should apply!

Kosta Derpanis

@CSProfKGD

5 years

I have MSc/PhD positions available to work with me on #computervision at @RyersonU for Fall 2019 supported by @NSERC_CRSNG . Email me your CV to apply! (Please mention this tweet in your email.) Please retweet widely

0

29

43

0

3

Antonio Loquercio

@antoniloq

2 years

@pabbeel @therobotbrains @geoffreyhinton - Do you think we are making progress towards A(G)I or just overfitting to some (biased) datasets? - How does the current progress compare to what you expected 10 years ago (after the "ImageNet" moment)?

0

3

Antonio Loquercio

@antoniloq

10 months

@AjdDavison @xiao_ted Respectfully disagree on this. Learning is definitely the biggest enabler. We got 90% of the performance without any bells and whistles (state estimation from an off-the-shelf realsense and a gate detector :))

1

0

3

Antonio Loquercio

@antoniloq

3 years

I feel very lucky to be able to pay for my insulin. I hope that more governments will follow and save the lives of less lucky diabetics people.

James Talarico

@jamestalarico

3 years

3 years ago I almost slipped into a coma and died. I was diagnosed with type 1 diabetes. I paid $684 for my first 30-day supply of insulin. Some pay $1000+ a month. Many die without it. Today I introduced legislation to cap the price of insulin at $50 a month. #txlege