Join The Townhall: Share Your Questions

The thrilling Scene Understanding for Autonomous Drone Challenge has officially concluded, and now it’s time to unveil the winning strategies that took participants to the top! :trophy:

:tada: We invite you to join our exclusive Towhall Event, where the challenge winners will share their insights, approaches, and techniques that led them to victory. Whether you are a seasoned participant or a passionate learner, this event promises to be a treasure trove of knowledge and inspiration!

:spiral_calendar: 12th August, 2023 :alarm_clock: 2:30 PM CET
:computer: Join the townhall on Zoom

:sparkles: What to Expect

  • Learn from the best: Discover the winning solutions from the brilliant minds behind the top-performing teams.

  • Collaborate and discuss: Engage with other participants to exchange ideas, improve your solutions, and explore new possibilities in AI.

  • Stay up-to-date: Get the latest updates about the challenge and future AIcrowd events.

:speech_balloon: Have Questions for the Winners?

If you’ve ever wondered about the secret sauce behind their success, now is your chance to get your burning questions answered directly from the challenge organizers and winners! Simply drop your questions and queries in the comments section below this post, and we’ll make sure to address them during the event.


This text will be hidden

:handshake: Meet Your Fellow Participants IRL

This is a fantastic opportunity to network and connect with other challenge participants from the AIcrowd community. Share experiences, discuss innovative ideas, and foster meaningful connections with like-minded individuals passionate about AI.

:raising_hand_woman: Got a burning question for the winners? Get it answered by challenge organisers and challenge winners. :speech_balloon: Drop your questions and query before the event on this post.

:spiral_calendar: Mark your calendars, prepare your questions, and get ready to immerse yourself in an inspiring event! Let’s celebrate the success of the challenge and ignite new possibilities together.

Hope to see you soon!

Team AIcrowd

  • did you team used extra data ? if yes, how did you change / assess the categories differences between the competition data and the extra data (specifically grass / soil / gravel) ?
  • the competition dataset was small and black and white with repetitive set of images: did it bother you in setting up your solution ? what guided you towards your solution to overcome such small and unusual dataset ?
  • did you correct manually or using a model the categories in the competition dataset ? if manually, what tool did you use ?
  • the dataset is composed of set of images of different zoom: does your solution take the zoom into account or does the model takes the pictures independantly of the zoom ?
  • what are the characteristic of the gpu used ? what batch / image size are you using ?
  • if you used mmsegmentation, what version of architecture and backbone did you used ? if the backbones weren’t available in the mmsegmentation, how did you hear about the particular backbones you used outside of mmsegmentation ? what guided you towards picking this backbones given the extensive number of publications nowadays ?
  • the backbones included in mmsegmentation were usually overfitting, what were your solution ?
  • did you used any “tricks” to reduce or augment the competition dataset ?
  • how did you build the validation dataset ? did you do anything special to build the validation set ? was the metric on validation set matching the public leaderboard ?
  • if you had more time for prediction, what would you have included in your solution ?
1 Like