• Corpus ID: 207915315

Detecting Waterborne Debris with Sim2Real and Randomization

@inproceedings{Fu2019DetectingWD,
  title={Detecting Waterborne Debris with Sim2Real and Randomization},
  author={Jie Fu and Ritchie Ng and Mirgahney Mohamed and Yi Tay and Kris Sankaran and Shangbang Long and A. Canziani and Christopher Joseph Pal and Moustapha Ciss{\'e}},
  year={2019}
}
From palpable marine debris to microplastics, marine debris pollution has been a perennial problem. In recent years, there is an emergence of largescale clean-up efforts making its way around the world. Complimentary to large-scale clean-up efforts, there is a nascent area in the use of unmanned and remote vehicles for detecting and removing debris. In this project, our focus is on marine debris detection where we propose to train a waterborne debris detector based on a mixture of real and… 
aiforsocialgood.github.io

Figures from this paper

References

SHOWING 1-8 OF 8 REFERENCES
Synthetic Examples Improve Generalization for Rare Classes
TLDR
The experiments reveal that synthetic data can considerably reduce error rates for classes that are rare, that as the amount of simulated data is increased, accuracy on the target class improves, and that high variation of simulation data provides maximum performance gain.
Classification of Trash for Recyclability Status
TLDR
The objective of this project is to take images of a single piece of recycling or garbage and classify it into six classes consisting of glass, paper, metal, plastic, cardboard, and trash, and create a dataset that contains around 400-500 images for each class.
Training Deep Networks with Synthetic Data: Bridging the Reality Gap by Domain Randomization
TLDR
This work presents a system for training deep neural networks for object detection using synthetic images that relies upon the technique of domain randomization, in which the parameters of the simulator are randomized in non-realistic ways to force the neural network to learn the essential features of the object of interest.
Domain randomization for transferring deep neural networks from simulation to the real world
TLDR
This paper explores domain randomization, a simple technique for training models on simulated images that transfer to real images by randomizing rendering in the simulator, and achieves the first successful transfer of a deep neural network trained only on simulated RGB images to the real world for the purpose of robotic control.
End-to-End Active Object Tracking and Its Real-World Deployment via Reinforcement Learning
TLDR
The tracker trained in simulators (ViZDoom and Unreal Engine) demonstrates good generalization behaviors in the case of unseen object moving paths, unseen object appearances, unseen backgrounds, and distracting objects and can restore tracking after occasional lost of the target being tracked.
YOLOv3: An Incremental Improvement
We present some updates to YOLO! We made a bunch of little design changes to make it better. We also trained this new network that's pretty swell. It's a little bigger than last time but more…
UnrealCV: Connecting Computer Vision to Unreal Engine
TLDR
An open-source plugin UnrealCV is created for a popular game engine Unreal Engine 4 (UE4) to enable researchers to build on these resources to create virtual worlds, provided they can access and modify the internal data structures of the games.
The ocean cleanup, May 2019
  • URL https: //www.theoceancleanup.com/
  • 2019