Pre-trained models ------------------ The currently available model checkpoints can be seen by running the command:: echofilter --list-checkpoints All current checkpoints were trained on data acquired by `FORCE `__. Training Datasets ~~~~~~~~~~~~~~~~~ Stationary ^^^^^^^^^^ :data collection: bottom-mounted :term:`stationary`, autonomous :orientation: uplooking :echosounder: 120 kHz Simrad WBAT :locations: - FORCE tidal power demonstration site, Minas Passage - 45°21'47.34"N 64°25'38.94"W - December 2017 through November 2018 - SMEC, Grand Passage - 44°15'49.80"N 66°20'12.60"W - December 2019 through January 2020 :organization: FORCE Mobile ^^^^^^ :data collection: vessel-based 24-hour transect surveys :orientation: downlooking :echosounder: 120 kHz Simrad EK80 :locations: - FORCE tidal power demonstration site, Minas Passage - 45°21'57.58"N 64°25'50.97"W - May 2016 through October 2018 :organization: FORCE .. _Model checkpoints: Model checkpoints ~~~~~~~~~~~~~~~~~ The architecture used for all current models is a U-Net with a backbone of 6 EfficientNet blocks in each direction (encoding and decoding). There are horizontal skip connections between compression and expansion blocks at the same spatial scale and a latent space of 32 channels throughout the network. The depth dimension of the input is halved (doubled) after each block, whilst the time dimension is halved (doubled) every other block. Details for notable model checkpoints are provided below. :conditional_mobile-stationary2_effunet6x2-1_lc32_v2.2: - Trained on both :term:`upfacing` :term:`stationary` and :term:`downfacing` :term:`mobile` data. - Jaccard Index of **96.84%** on :term:`downfacing` :term:`mobile` and **94.51%** on :term:`upfacing` :term:`stationary` :term:`validation` data. - Default model checkpoint. :conditional_mobile-stationary2_effunet6x2-1_lc32_v2.1: - Trained on both :term:`upfacing` :term:`stationary` and :term:`downfacing` :term:`mobile` data. - Jaccard Index of 96.8% on :term:`downfacing` :term:`mobile` and 94.4% on :term:`upfacing` :term:`stationary` :term:`validation` data. :conditional_mobile-stationary2_effunet6x2-1_lc32_v2.0: - Trained on both :term:`upfacing` :term:`stationary` and :term:`downfacing` :term:`mobile` data. - Jaccard Index of 96.62% on :term:`downfacing` :term:`mobile` and 94.29% on :term:`upfacing` :term:`stationary` :term:`validation` data. - :term:`Sample` outputs on :term:`upfacing` :term:`stationary` data were thoroughly verified via manual inspection by trained analysts. :stationary2_effunet6x2-1_lc32_v2.1: - Trained on :term:`upfacing` :term:`stationary` data only. - Jaccard Index of 94.4% on :term:`upfacing` :term:`stationary` :term:`validation` data. :stationary2_effunet6x2-1_lc32_v2.0: - Trained on :term:`upfacing` :term:`stationary` data only. - Jaccard Index of 94.41% on :term:`upfacing` :term:`stationary` :term:`validation` data. - :term:`Sample` outputs thoroughly were thoroughly verified via manual inspection by trained analysts. :mobile_effunet6x2-1_lc32_v1.0: - Trained on :term:`downfacing` :term:`mobile` data only. .. raw:: latex \clearpage