{ "version": 1, "source_sha256": "92e118e74a8eaf0d253122a3aab4728ec2984d94a0210b618489f4a8ca052aa0", "datasets": [ { "dataset": "moin", "title": "Vegetation", "images": 10, "methods": { "whole": { "name": "DINOv3", "mean": 0.0610436264096801, "sample_sd": 0.02606464087243846, "n": 10, "median": 0.07039481847891395, "scores": [ { "image_id": "moin_01", "f1": 0.08207315644523339 }, { "image_id": "moin_02", "f1": 0.03682463160239658 }, { "image_id": "moin_03", "f1": 0.03909498979788084 }, { "image_id": "moin_04", "f1": 0.08215612225781334 }, { "image_id": "moin_05", "f1": 0.06369117250055055 }, { "image_id": "moin_06", "f1": 0.0774048945348962 }, { "image_id": "moin_07", "f1": 0.09762514724895907 }, { "image_id": "moin_08", "f1": 0.026479019597973632 }, { "image_id": "moin_09", "f1": 0.07709846445727735 }, { "image_id": "moin_10", "f1": 0.02798866565382011 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "sam3": { "name": "SAM3", "mean": 0.02023451650834254, "sample_sd": 0.015701529299251785, "n": 10, "median": 0.019046747858618686, "scores": [ { "image_id": "moin_01", "f1": 0.02178335984847502 }, { "image_id": "moin_02", "f1": 0.01631013586876235 }, { "image_id": "moin_03", "f1": 0.007507288611952792 }, { "image_id": "moin_04", "f1": 0.03810355332933491 }, { "image_id": "moin_05", "f1": 0.02630638581750314 }, { "image_id": "moin_06", "f1": 0.004525996435755741 }, { "image_id": "moin_07", "f1": 0.006408098030983862 }, { "image_id": "moin_08", "f1": 0.00675654829371639 }, { "image_id": "moin_09", "f1": 0.05280333557170674 }, { "image_id": "moin_10", "f1": 0.021840463275234444 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s05": { "name": "Our technology · 5%", "mean": 0.5956753352704712, "sample_sd": 0.10626377259733716, "n": 10, "median": 0.6151140536347952, "scores": [ { "image_id": "moin_01", "f1": 0.5694581065572494 }, { "image_id": "moin_02", "f1": 0.5960330127459696 }, { "image_id": "moin_03", "f1": 0.7177595713692377 }, { "image_id": "moin_04", "f1": 0.5101425842487705 }, { "image_id": "moin_05", "f1": 0.6892456492804568 }, { "image_id": "moin_06", "f1": 0.7179274330317088 }, { "image_id": "moin_07", "f1": 0.4507799484404775 }, { "image_id": "moin_08", "f1": 0.6481313983811287 }, { "image_id": "moin_09", "f1": 0.4230805541260928 }, { "image_id": "moin_10", "f1": 0.6341950945236208 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s10": { "name": "Our technology · 10%", "mean": 0.4414449403773494, "sample_sd": 0.08262872857592182, "n": 10, "median": 0.43741665040875927, "scores": [ { "image_id": "moin_01", "f1": 0.47276376723389335 }, { "image_id": "moin_02", "f1": 0.39022308248418636 }, { "image_id": "moin_03", "f1": 0.5568510787188672 }, { "image_id": "moin_04", "f1": 0.4020695335836252 }, { "image_id": "moin_05", "f1": 0.5080047772818979 }, { "image_id": "moin_06", "f1": 0.5516406616998831 }, { "image_id": "moin_07", "f1": 0.3512337131506934 }, { "image_id": "moin_08", "f1": 0.47587362695994745 }, { "image_id": "moin_09", "f1": 0.3264846119336977 }, { "image_id": "moin_10", "f1": 0.37930455072680214 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 } }, "paired_comparisons": { "s05_vs_whole": { "mean_difference": 0.5346317088607911, "wins": 10, "ties": 0, "n": 10 }, "s05_vs_sam3": { "mean_difference": 0.5754408187621287, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_whole": { "mean_difference": 0.3804013139676693, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_sam3": { "mean_difference": 0.42121042386900687, "wins": 10, "ties": 0, "n": 10 } } }, { "dataset": "fungal_network", "title": "Fungal networks", "images": 10, "methods": { "whole": { "name": "DINOv3", "mean": 0.07625523151460245, "sample_sd": 0.057430314971912456, "n": 10, "median": 0.0441918433419967, "scores": [ { "image_id": "fungal_network_01", "f1": 0.03923696515514304 }, { "image_id": "fungal_network_02", "f1": 0.041606702914250684 }, { "image_id": "fungal_network_03", "f1": 0.046639250899090864 }, { "image_id": "fungal_network_04", "f1": 0.031135826824818157 }, { "image_id": "fungal_network_05", "f1": 0.13498431356391638 }, { "image_id": "fungal_network_06", "f1": 0.13878042757975761 }, { "image_id": "fungal_network_07", "f1": 0.19365478371531944 }, { "image_id": "fungal_network_08", "f1": 0.041744435784902535 }, { "image_id": "fungal_network_09", "f1": 0.037353057119788875 }, { "image_id": "fungal_network_10", "f1": 0.057416551589036934 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "sam3": { "name": "SAM3", "mean": 0.03350922467311785, "sample_sd": 0.02716833569115167, "n": 10, "median": 0.028779826849129386, "scores": [ { "image_id": "fungal_network_01", "f1": 0.03594759938792127 }, { "image_id": "fungal_network_02", "f1": 0.03175137460637952 }, { "image_id": "fungal_network_03", "f1": 0.025808279091879247 }, { "image_id": "fungal_network_04", "f1": 0.011571975828405301 }, { "image_id": "fungal_network_05", "f1": 0.021533019607697164 }, { "image_id": "fungal_network_06", "f1": 0.03613684057602323 }, { "image_id": "fungal_network_07", "f1": 0.021093357227280802 }, { "image_id": "fungal_network_08", "f1": 0.0 }, { "image_id": "fungal_network_09", "f1": 0.0992751329061589 }, { "image_id": "fungal_network_10", "f1": 0.051974667499433046 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s05": { "name": "Our technology · 5%", "mean": 0.29015282699332084, "sample_sd": 0.10775473782156252, "n": 10, "median": 0.2678884110248383, "scores": [ { "image_id": "fungal_network_01", "f1": 0.4141210788357951 }, { "image_id": "fungal_network_02", "f1": 0.39675198653017657 }, { "image_id": "fungal_network_03", "f1": 0.23633707185775799 }, { "image_id": "fungal_network_04", "f1": 0.44284741160285274 }, { "image_id": "fungal_network_05", "f1": 0.3613111578085646 }, { "image_id": "fungal_network_06", "f1": 0.26879040871700777 }, { "image_id": "fungal_network_07", "f1": 0.2669864133326688 }, { "image_id": "fungal_network_08", "f1": 0.20433250426366648 }, { "image_id": "fungal_network_09", "f1": 0.12660149210138236 }, { "image_id": "fungal_network_10", "f1": 0.18344874488333665 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s10": { "name": "Our technology · 10%", "mean": 0.24364695723118274, "sample_sd": 0.09303509731797634, "n": 10, "median": 0.2347511175736393, "scores": [ { "image_id": "fungal_network_01", "f1": 0.3147056174116502 }, { "image_id": "fungal_network_02", "f1": 0.27390593556101256 }, { "image_id": "fungal_network_03", "f1": 0.18357401522472602 }, { "image_id": "fungal_network_04", "f1": 0.35480039419236814 }, { "image_id": "fungal_network_05", "f1": 0.4008284418748101 }, { "image_id": "fungal_network_06", "f1": 0.2286518909909146 }, { "image_id": "fungal_network_07", "f1": 0.240850344156364 }, { "image_id": "fungal_network_08", "f1": 0.18241594078240284 }, { "image_id": "fungal_network_09", "f1": 0.10959613121084782 }, { "image_id": "fungal_network_10", "f1": 0.14714086090673092 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 } }, "paired_comparisons": { "s05_vs_whole": { "mean_difference": 0.21389759547871842, "wins": 10, "ties": 0, "n": 10 }, "s05_vs_sam3": { "mean_difference": 0.25664360232020306, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_whole": { "mean_difference": 0.1673917257165803, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_sam3": { "mean_difference": 0.2101377325580649, "wins": 10, "ties": 0, "n": 10 } } }, { "dataset": "neon", "title": "Aerial ecosystems", "images": 10, "methods": { "whole": { "name": "DINOv3", "mean": 0.03360191510580875, "sample_sd": 0.010911652324640654, "n": 10, "median": 0.03083712143076272, "scores": [ { "image_id": "neon_01", "f1": 0.03515453333828102 }, { "image_id": "neon_02", "f1": 0.030573374919883725 }, { "image_id": "neon_03", "f1": 0.022225522378974932 }, { "image_id": "neon_04", "f1": 0.02637827166516797 }, { "image_id": "neon_05", "f1": 0.02806492528343843 }, { "image_id": "neon_06", "f1": 0.05726713527592723 }, { "image_id": "neon_07", "f1": 0.04687499548783981 }, { "image_id": "neon_08", "f1": 0.02343101217229743 }, { "image_id": "neon_09", "f1": 0.031100867941641713 }, { "image_id": "neon_10", "f1": 0.034948512594635196 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "sam3": { "name": "SAM3", "mean": 0.040634268380721976, "sample_sd": 0.04445752919187413, "n": 10, "median": 0.023582281795223183, "scores": [ { "image_id": "neon_01", "f1": 0.012704689767286025 }, { "image_id": "neon_02", "f1": 0.041770635858206076 }, { "image_id": "neon_03", "f1": 0.009562559932318044 }, { "image_id": "neon_04", "f1": 0.002007593041141287 }, { "image_id": "neon_05", "f1": 0.048032212929391305 }, { "image_id": "neon_06", "f1": 0.1505617879610194 }, { "image_id": "neon_07", "f1": 0.01798616883710839 }, { "image_id": "neon_08", "f1": 0.07655247189030288 }, { "image_id": "neon_09", "f1": 0.027190599714841675 }, { "image_id": "neon_10", "f1": 0.01997396387560469 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s05": { "name": "Our technology · 5%", "mean": 0.5492423518488275, "sample_sd": 0.13567393842789238, "n": 10, "median": 0.5809433197921461, "scores": [ { "image_id": "neon_01", "f1": 0.6073031873107452 }, { "image_id": "neon_02", "f1": 0.522649562914274 }, { "image_id": "neon_03", "f1": 0.7035850965410275 }, { "image_id": "neon_04", "f1": 0.6450909693799466 }, { "image_id": "neon_05", "f1": 0.27895901669233025 }, { "image_id": "neon_06", "f1": 0.6783588311964389 }, { "image_id": "neon_07", "f1": 0.554583452273547 }, { "image_id": "neon_08", "f1": 0.3759739395249021 }, { "image_id": "neon_09", "f1": 0.4980408157780168 }, { "image_id": "neon_10", "f1": 0.6278786468770458 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s10": { "name": "Our technology · 10%", "mean": 0.35982512016680956, "sample_sd": 0.13457562702246711, "n": 10, "median": 0.3855815886725114, "scores": [ { "image_id": "neon_01", "f1": 0.4012717725090851 }, { "image_id": "neon_02", "f1": 0.3747228357685551 }, { "image_id": "neon_03", "f1": 0.5855847727572825 }, { "image_id": "neon_04", "f1": 0.39644034157646774 }, { "image_id": "neon_05", "f1": 0.20321329753270212 }, { "image_id": "neon_06", "f1": 0.49131727125750757 }, { "image_id": "neon_07", "f1": 0.3573392315217963 }, { "image_id": "neon_08", "f1": 0.26407845604638536 }, { "image_id": "neon_09", "f1": 0.12308704563915826 }, { "image_id": "neon_10", "f1": 0.4011961770591564 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 } }, "paired_comparisons": { "s05_vs_whole": { "mean_difference": 0.5156404367430187, "wins": 10, "ties": 0, "n": 10 }, "s05_vs_sam3": { "mean_difference": 0.5086080834681053, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_whole": { "mean_difference": 0.3262232050610009, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_sam3": { "mean_difference": 0.3191908517860877, "wins": 10, "ties": 0, "n": 10 } } }, { "dataset": "coralscapes", "title": "Coralscapes", "images": 10, "methods": { "whole": { "name": "DINOv3", "mean": 0.04099941675087438, "sample_sd": 0.009724220016379712, "n": 10, "median": 0.038006856504990624, "scores": [ { "image_id": "coralscapes_01", "f1": 0.05136895237377318 }, { "image_id": "coralscapes_02", "f1": 0.038100831897927186 }, { "image_id": "coralscapes_03", "f1": 0.05360510010744393 }, { "image_id": "coralscapes_04", "f1": 0.030537971996277453 }, { "image_id": "coralscapes_05", "f1": 0.037557358952178814 }, { "image_id": "coralscapes_06", "f1": 0.03791288111205406 }, { "image_id": "coralscapes_07", "f1": 0.04086282765253009 }, { "image_id": "coralscapes_08", "f1": 0.026697277731694575 }, { "image_id": "coralscapes_09", "f1": 0.03728425826513959 }, { "image_id": "coralscapes_10", "f1": 0.05606670741972491 } ], "annotated_boundary_mean": 0.0215254768813522, "annotated_boundary_n": 10 }, "sam3": { "name": "SAM3", "mean": 0.03614715239647749, "sample_sd": 0.023780088030356313, "n": 10, "median": 0.03207108356469571, "scores": [ { "image_id": "coralscapes_01", "f1": 0.06576161462757618 }, { "image_id": "coralscapes_02", "f1": 0.01779160098056162 }, { "image_id": "coralscapes_03", "f1": 0.060566152085020836 }, { "image_id": "coralscapes_04", "f1": 0.005063577062622655 }, { "image_id": "coralscapes_05", "f1": 0.03301777173463232 }, { "image_id": "coralscapes_06", "f1": 0.04021172802517151 }, { "image_id": "coralscapes_07", "f1": 0.030231531483124727 }, { "image_id": "coralscapes_08", "f1": 0.005596124371434076 }, { "image_id": "coralscapes_09", "f1": 0.0311243953947591 }, { "image_id": "coralscapes_10", "f1": 0.07210702819987189 } ], "annotated_boundary_mean": 0.11261083310569978, "annotated_boundary_n": 10 }, "s05": { "name": "Our technology · 5%", "mean": 0.43618629140653525, "sample_sd": 0.1363145862227491, "n": 10, "median": 0.438120871686717, "scores": [ { "image_id": "coralscapes_01", "f1": 0.24647834987905756 }, { "image_id": "coralscapes_02", "f1": 0.404438789309274 }, { "image_id": "coralscapes_03", "f1": 0.22781948141468092 }, { "image_id": "coralscapes_04", "f1": 0.4718029540641599 }, { "image_id": "coralscapes_05", "f1": 0.3932187250627826 }, { "image_id": "coralscapes_06", "f1": 0.522673806234811 }, { "image_id": "coralscapes_07", "f1": 0.5813076089707173 }, { "image_id": "coralscapes_08", "f1": 0.6608518474055758 }, { "image_id": "coralscapes_09", "f1": 0.4741181845548479 }, { "image_id": "coralscapes_10", "f1": 0.37915316716944486 } ], "annotated_boundary_mean": 0.03768124393727508, "annotated_boundary_n": 10 }, "s10": { "name": "Our technology · 10%", "mean": 0.26594125540804014, "sample_sd": 0.07435986304707257, "n": 10, "median": 0.24598196232021607, "scores": [ { "image_id": "coralscapes_01", "f1": 0.16693805669394407 }, { "image_id": "coralscapes_02", "f1": 0.30679798216696863 }, { "image_id": "coralscapes_03", "f1": 0.1960522395930811 }, { "image_id": "coralscapes_04", "f1": 0.22051466306610146 }, { "image_id": "coralscapes_05", "f1": 0.3303421873148029 }, { "image_id": "coralscapes_06", "f1": 0.2563471884756811 }, { "image_id": "coralscapes_07", "f1": 0.2334136689680709 }, { "image_id": "coralscapes_08", "f1": 0.42248225326823713 }, { "image_id": "coralscapes_09", "f1": 0.23561673616475107 }, { "image_id": "coralscapes_10", "f1": 0.2909075783687629 } ], "annotated_boundary_mean": 0.04717855509167597, "annotated_boundary_n": 10 } }, "paired_comparisons": { "s05_vs_whole": { "mean_difference": 0.39518687465566077, "wins": 10, "ties": 0, "n": 10 }, "s05_vs_sam3": { "mean_difference": 0.4000391390100577, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_whole": { "mean_difference": 0.22494183865716572, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_sam3": { "mean_difference": 0.22979410301156267, "wins": 10, "ties": 0, "n": 10 } } }, { "dataset": "pmid", "title": "Aquatic microscopy", "images": 10, "methods": { "whole": { "name": "DINOv3", "mean": 0.028068520475573854, "sample_sd": 0.0073911995894940085, "n": 10, "median": 0.028101322130057407, "scores": [ { "image_id": "pmid_01", "f1": 0.024427939920495786 }, { "image_id": "pmid_02", "f1": 0.02715956653727992 }, { "image_id": "pmid_03", "f1": 0.03213030493948211 }, { "image_id": "pmid_04", "f1": 0.03162902463171502 }, { "image_id": "pmid_05", "f1": 0.023764146421987306 }, { "image_id": "pmid_06", "f1": 0.01852718668321354 }, { "image_id": "pmid_07", "f1": 0.029043077722834897 }, { "image_id": "pmid_08", "f1": 0.0443840532936276 }, { "image_id": "pmid_09", "f1": 0.029833177504373837 }, { "image_id": "pmid_10", "f1": 0.019786727100728464 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "sam3": { "name": "SAM3", "mean": 0.0903737389148302, "sample_sd": 0.031691730347725676, "n": 10, "median": 0.07773808770380888, "scores": [ { "image_id": "pmid_01", "f1": 0.07344830651164713 }, { "image_id": "pmid_02", "f1": 0.046990380259832154 }, { "image_id": "pmid_03", "f1": 0.0689773605057349 }, { "image_id": "pmid_04", "f1": 0.1226196316876787 }, { "image_id": "pmid_05", "f1": 0.062300823687330774 }, { "image_id": "pmid_06", "f1": 0.07956332696856397 }, { "image_id": "pmid_07", "f1": 0.14513606682809777 }, { "image_id": "pmid_08", "f1": 0.12167933922927718 }, { "image_id": "pmid_09", "f1": 0.10710930503108572 }, { "image_id": "pmid_10", "f1": 0.07591284843905378 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s05": { "name": "Our technology · 5%", "mean": 0.4555654211073564, "sample_sd": 0.11847755518021147, "n": 10, "median": 0.4927078720099989, "scores": [ { "image_id": "pmid_01", "f1": 0.19171244037154797 }, { "image_id": "pmid_02", "f1": 0.4985030379871189 }, { "image_id": "pmid_03", "f1": 0.4732840405077838 }, { "image_id": "pmid_04", "f1": 0.5228170056328157 }, { "image_id": "pmid_05", "f1": 0.5969344115575956 }, { "image_id": "pmid_06", "f1": 0.33109126398951516 }, { "image_id": "pmid_07", "f1": 0.4869127060328788 }, { "image_id": "pmid_08", "f1": 0.4000311721370357 }, { "image_id": "pmid_09", "f1": 0.5378949768676985 }, { "image_id": "pmid_10", "f1": 0.5164731559895744 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 }, "s10": { "name": "Our technology · 10%", "mean": 0.28821301820223283, "sample_sd": 0.13816206494399677, "n": 10, "median": 0.3683619223311319, "scores": [ { "image_id": "pmid_01", "f1": 0.058239872009496194 }, { "image_id": "pmid_02", "f1": 0.4101460626868934 }, { "image_id": "pmid_03", "f1": 0.1958256356144527 }, { "image_id": "pmid_04", "f1": 0.35995403536985543 }, { "image_id": "pmid_05", "f1": 0.42081280985657143 }, { "image_id": "pmid_06", "f1": 0.1197240189962344 }, { "image_id": "pmid_07", "f1": 0.3767698092924084 }, { "image_id": "pmid_08", "f1": 0.16167504335721475 }, { "image_id": "pmid_09", "f1": 0.38623291877944005 }, { "image_id": "pmid_10", "f1": 0.39274997605976125 } ], "annotated_boundary_mean": null, "annotated_boundary_n": 0 } }, "paired_comparisons": { "s05_vs_whole": { "mean_difference": 0.4274969006317826, "wins": 10, "ties": 0, "n": 10 }, "s05_vs_sam3": { "mean_difference": 0.36519168219252623, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_whole": { "mean_difference": 0.260144497726659, "wins": 10, "ties": 0, "n": 10 }, "s10_vs_sam3": { "mean_difference": 0.19783927928740258, "wins": 9, "ties": 0, "n": 10 } } } ], "images": 50, "methods": { "whole": "DINOv3", "sam3": "SAM3", "s05": "Our technology · 5%", "s10": "Our technology · 10%" }, "primary_metric": "RGB-edge F1: an appearance proxy, not biological object or species accuracy.", "metric_calculation": "Grayscale luminance 0.2126R+0.7152G+0.0722B; Gaussian sigma=1; Sobel gradients; strongest15% eligible gradients with minimum threshold1e-6. Predicted boundaries are changes in neighboring partition labels. Precision and recall match edges within Manhattan distance2 working pixels. F1=2PR/(P+R). Two-pixel image margins excluded.", "clustering": "Fixed KMeans K=6 for DINOv3 and local5/10, using16 unwhitened PCA axes; 8192 same fitting locations,n_init10,seed42. Separate fitted centers per method. This benchmark predates automatic K selection and does not score GMM lambda4.", "local_treatment": "Crop enlargement384x384,four inverse-aligned flips,approximately75% overlap,Hann blending,noAnyUp.", "sam3": "HuggingFace automatic MaskGenerationPipeline/Sam3TrackerModel; zero user-supplied prompts; highest predictedIoU overlap resolution; threshold0.88 and stability0.95.", "error_bars": "Sample standard deviation across ten images, not confidence intervals. Equal weight per photograph.", "conclusion": "Both5% and10% have higher mean RGB-edge F1 than DINOv3 and SAM3 in each of five domains. This conclusion is specific to these sampled photographs and this appearance metric.", "annotation_caveat": "Only Coralscapes has eligible annotated pixel class boundaries here. Its mean boundaryF1 is higher for SAM3 than for either local focus scale. Box annotations are not used as pixel masks; missing references are not zero scores.", "inference_reused": true }