Fig. 3From: Comprehensive benchmark and architectural analysis of deep learning models for nanopore sequencing basecallingBenchmark of architecture components. a Top 25 best performing model combinations. b Comparison of CTC to CRF decoder. c Comparison of simple (Bonito, CATCAller, SACall) to complex (Causalcall, Mincall, URNano) convolutions. d Comparison of bidirection LSTM depth (1, 3, or 5 layers). e Comparison of RNN to Transformer encodersBack to article page