Below are the highest classification accuracies achieved on the Dex-Net 2.0 dataset on a randomized 80-20 train-validation split using various splitting rules:
The current leader is a ConvNet submitted by nomagic.ai. GQ is our best GQ-CNN for Dex-Net 2.0.
We believe grasping performance on the physical robot can be improved if these validation error rates can be further reduced by modifications to the network architecture and optimization. If you achieve superior numbers on a randomized validation set, please email Jeff Mahler (email@example.com) with the subject “Dex-Net 2.0 Benchmark Submission” and we will consider testing on our ABB YuMi.