0 Follower
0 Following
SKP0
Location
BE
Badges
2
0
0
Activity
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Mon
Wed
Fri
Ratings Progression
Loading...
Challenge Categories
Loading...
Challenges Entered
Completed
MABe Team
AIcrowd
Behavioral Representation Learning from Animal Poses.
Latest submissions
See All| graded | 197891 | ||
| graded | 197881 | ||
| graded | 197862 |
Parametrix.ai
MIT
THU_SIGS
Latest submissions
See All| graded | 181283 | ||
| graded | 181273 | ||
| graded | 181217 |
Robustness and teamwork in a massively multiagent environment
Latest submissions
Round 2 - Active | Claim AWS Credits by beating the baseline
Latest submissions
See All| graded | 197891 | ||
| graded | 197881 | ||
| graded | 197862 |
Round 2 - Active | Claim AWS Credits by beating the baseline
Latest submissions
See All| graded | 197849 | ||
| graded | 197846 | ||
| graded | 197845 |
| Participant | Rating |
|---|
| Participant | Rating |
|---|
SKP0 has not joined any teams yet...
Multi Agent Behavior Challenge 2022
[Round-2 Update] $400 AWS Credits Per Team - How To Win & Claim Them
Over 2 years agoHi, I beat the baselines on both tasks.
Ant & Beetle
#183770
Mean F1 Score: 0.557 → 0.580
Mouse Triplet
#183189
Mean F1 Score: 0.217 → 0.230
I am a PhD student with a broad interest in machine learning, including computer vision tasks.
SKP0 has not provided any information yet.
Share your solutions!
Over 2 years agoMy solution was made early on in the competition and uses only video data. It held the first place initially, though got surpassed later on. In hindsight I think I could have benefitted from using the keypoints as well.
I used an ensemble of pre-trained vision models by concatenating the output features of the vision models (resnet18 and MobileNetV3-Small). This results in a large vector which is then reduced to size 64 by PCA for the beetles challenge, which is the final embedding. For the mouse part of the challenge I instead reduce the size down to 32 in a similar way, and then concatenate to this the difference of the feature vector from 40 frames in the past and 40 frames in the future (window size of 80 as in the solution of @edhayes1) to also have dynamic information present in the embedding.