Skip to main content
. Author manuscript; available in PMC: 2020 Dec 17.
Published in final edited form as: Comput Vis ECCV. 2020 Dec 4;12363:103–120. doi: 10.1007/978-3-030-58523-5_7

Table 1.

Active learning performance comparison on our EM-R50 connectomics benchmark. Our two-stream query suggestion approach significantly out-perform previous methods in terms of the ROI proposal accuracy (higher is better).

Method Synapse
 Mitochondria
Round 1 Round 2 Round 1 Round 2
Random 0.824 0.871 0.704 0.749
Core-Set [43] 0.847 0.895 0.726 0.767
Learning-Loss1 [53] 0.832 0.889 0.724 0.771
Two-Stream (Ours) 0.892 0.926 0.802 0.809