ADAPT Cascade Competition

Sim2Real object Detection and Pose-Estimation Challenge

We were pleased to invite participants to the ADAPT Cascade Competition, dedicated to benchmarking and testing machine learning algorithms and solutions for typical industrial challenges.

In this Cascade competition, simulation data was used across two main benchmarks to test and evaluate competitors' solutions for Object Detection and Classification, as well as Pose Estimation of typical industrial parts. Competition took place on Codalab.

Representative images

Results

The Sim2Real Object Detection task yielded an average precision of 0.70. This outcome stems from real-image datasets featuring 3D printed object parts, staged with added complexity through background clutter. Despite these challenges, we anticipated that a well-trained, deep-learning-based detector would overcome the obstacles.

User	Total Submissions	Successful	Failed
Giwan	11	11	0
stefiOns	18	13	5
Total	29	24	5

Awards

Please be informed that the winner of each of the two ongoing ADAPT Sim2Real competitions for Object Detection and Pose Estimations competition will be awarded with a “CEA-Moonshot Prize” to attend the ICRA-2023 conference in the UK during 29 May to 2 June 2023.

The Prizes, limited to 1500€ per winning team (or individual) will be paid in the form of reimbursement of the costs of travel, subsistence and conference registration fees upon submission of receipts along with a travel expense claim form.  A certificate will also be provided to the awardees during ICRA.

N.B.  The next METRICS/ADAPT Field competition will occur during ICRA as well, and the winner of that competition will also receive a 1000 € Prize in the form of travel cost reimbursement.