Table 3.

Accuracy of the model without post-processing, and with post processing according to our goals, on the cross-validation, with thresholds tuned on T1 and applied on T2.

Species	Raw Accuracy	G1 Accuracy	G2 Accuracy	G3 Accuracy
Abudefduf vaigiensis	0.51	0.61	0.92	0.97
Acanthurus leucosternon	0.61	0.7	0.92	0.94
Acanthurus lineatus	0.87	0.91	0.95	0.97
Amblyglyphidodon indicus	0.08	0.72	0.97	0.97
Chaetodon auriga	0.95	0.99	0.95	1
Chaetodon guttatissimus	0.16	0.88	0.72	0.96
Chaetodon trifascialis	0.97	0.9	0.96	0.98
Chaetodon trifasciatus	0.56	0.62	0.43	0.85
Chromis opercularis	0.68	0.83	0.03	1
Chromis ternatensis	0.01	0.47	0.97	0.87
Gomphosus caeruleus	0.24	0.31	0.89	0.75
Halichoeres hortulanus	0.51	0.57	1	0.9
Monotaxis grandoculis	0.77	0.82	0.99	0.98
Naso brevirostris	0.02	0.92	0.89	1
Naso elegans	0.89	0.91	0.99	0.97
Oxymonacanthus longirostris	0.36	0.46	0.72	0.8
Pomacentrus sulfureus	0.52	0.92	0.71	0.91
Thalassoma hardwicke	0.78	0.94	0.77	0.94
Zanclus cornutus	0.55	0.64	0.4	0.98
Zebrasoma scopas	0.61	0.66	0.99	0.8
Average	0.53	0.74	0.81	0.93
Standard deviation	0.30	0.19	0.25	0.07

Each line shows the result for a species, with: the species name, the accuracy of the model without post processing, and the accuracy of the model with post processing according to the 3 goals defined earlier.