| CARVIEW |
ALFRED Challenge
Leaderboard is now live for submissions to the ALFRED challenge! Humans have a success rate of 91% on unseen environments, but our baseline model has a 0.4% success rate. š¢
Challenge Winner: After 17 submissions, we saw a 10x increase in success rate to 4.5% by the challenge winners: Van-Quang Nguyen and Takayuki Okatani of Tohoku University!
Can you do even better? Code, precomputed features, and AI2Thor simulator are all available for a quick start on GitHub
Leaderboard
Embodied Vision, Actions & Language Workshop @ ECCV
Speakers and Panelists
 Kristen Grauman (Talk) |
 Nick Roy (Talk) |
 Chelsea Finn (Talk) |
 Jean Oh (Talk) |
 Jason Baldridge (Talk) |
 Abhinav Gupta (Talk) |
Live Sessions!
| There will be two live sessions including panels, poster presentations and best challenge submissions award: |
Aug 22 | |
| 7 - 7:45pm | Opening Remarks & Paper Talks (Recording) |
| 7:45 - 9pm | Poster Q&A |
Aug 23 | |
| 11am - 12pm: | Invited Speaker Panel (Recording) |
| 12 - 1pm: | Social Hour |
All times UTC-5 (EDT)
Poster Session
Workshop Details
The focus of this workshop is on embodied visual tasks that require the grounding of language to actions in real-world settings. Specifically, we want to draw focus to challenges like partial observability, continuous state spaces, and irrevocable actions for language-guided agents in visual environments. Such challenges are not captured by current datasets for grounding and embodiment [1, 2, 3].
- Key Topics
- Egocentric and Robotic vision
- Language Grounding
- Navigation and Motion Planning
- Interactive/Causal Reasoning
- Learning from Demonstration
- Task and Symbolic Planning
- Deep Reinforcement Learning
- Commonsense Reasoning
Submission Details
Contributed Papers
Standard ECCV 2020 format -- Submit papers to OpenReview
Challenge Papers
Participants are required to upload their model to our evaluation server. The evaluation server automatically evaluates the models on an unseen test set. Final numbers for the prize challenge will be frozen on Aug 5
Publication Options Archival vs Unofficial
Papers can submitted for publication in either the official proceedings (archival) or to be hosted on this website (unofficial). Both submission types can be presented at the workshop, but opting out of the proceedings allows you to submit your work for publication at another venue. Unofficial submissions are not required to be in the ECCV format. Please add either Archival or Unofficial as a keyword when submitting to indicate the correct submission track. No submissions will be made public during the review process. If there is any doubt we will contact authors to confirm the desired submission track.
Important Dates
| Contributed Papers | |
Challenge Papers | |
| | |
| | |
| | |
| | |
| Camera Ready | Sept 10 |
Abstract Deadline ECCV conference organizers have asked that there be a version of all papers (official and unofficial) in the conference system by Aug 5. We are interpreting this restriction as abstract only and that you will be allowed to make changes via OpenReview to the full paper through the deadline and all papers can be heavily revised for the camera ready. We are sorry for the new virtual restrictions. We will continue to update this space (including how to upload abstracts) as the conference organizers give us more information. Please don't hesitate to reach out with questions!
NEW! System Descriptions Submit your descriptions to OpenReview by Aug 10 (5pm Pacific Daylight Time).
Submission FAQ
- Do unofficial submissions need to be in ECCV format?
No - Can unofficial submissions already be published?
Yes - Page lengths for official submissions?
4 - 14 pages (ECCV rules require not major overlap with existing published work) - Page lengths for unofficial submissions?
Minimum 4 pages - Is reviewing blind?
Official are double blind, Unofficial are single - Supplemental Material
Can be merged to the end of the submission PDF - Competition Metric
Ranking and awards are based on Unseen Success Rate