|
|
SMURF: Self-Teaching Multi-Frame Unsupervised RAFT with
Full-Image Warping
Austin Stone*,
Daniel Mauerer*,
Alper Avaci,
Anelia Angelova,
Rico Jonschkowski
CVPR, 2021
Video
/
arXiv
/
Code
Improved the state of the art in unsupervised optical flow ~40% by
using RAFT architecture along with new ideas to improve the training signal.
|
|
The Distracting Control Suite – A Challenging Benchmark
for Reinforcement Learning from Pixels
Austin Stone*,
Oscar Ramirez,
Kurt Konolige,
Rico Jonschkowski*
arXiv, 2021
arXiv
/
Code
Created a new benchmark for vision-based control that includes variations in background, color, and camera pose. Showed that state of the art RL methods perform poorly on our benchmark,
highlighting the need for new methods which can cope with the visual complexities of the real world.
|
|
What Matters in Unsupervised Optical Flow
Rico Jonschkowski,
Austin Stone,
Jonathan T Barron,
Ariel Gordon,
Kurt Konolige,
Anelia Angelova
ECCV, 2020 (oral)
Video
/
arXiv
/
Code
Conducted a large scale study of many proposed innovations in Unsupervised Optical Flow
in order to determine what really matters.
|
|
Learning Object-conditioned Exploration Using Distributed Soft Actor Critic
Ayzaan Wahid,
Austin Stone,
Brian Ichter,
Kevin Chen,
Alexander Toshev
CoRL, 2020
arXiv
Trained a robot agent to navigate to objects from video using reinforcement learning.
|
|
Towards Object Detection from Motion
Rico Jonschkowski,
Austin Stone
arXiv, 2019
Video
/
arXiv
Developed a technique for weakly supervised object detection requiring only pairs of videos, one containing the object and the other not.
|
|
Teaching Compositionality to CNNs
Austin Stone,
Huayan Wang,
Michael Stark,
Yi Liu,
D. Scott Phoenix,
Dileep George
CVPR, 2017
arXiv
Showed that CNNs have very "uncompositional" feature activations, and applying the inductive bias
of compositionality improves performance in some settings.
|
|
Source
|
