--- language: - eng license: cc0-1.0 tags: - multilabel-image-classification - multilabel - generated_from_trainer base_model: BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze model-index: - name: BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze results: [] --- BeitEAU is a fine-tuned version of [BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze](https://huggingface.co/BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze). It achieves the following results on the test set: - Loss: 0.1647 - F1 Micro: 0.7446 - F1 Macro: 0.6015 - Accuracy: 0.2171 | Class | F1 per class | |----------|-------| | Acropore_branched | 0.7982 | | Acropore_digitised | 0.4713 | | Acropore_sub_massive | 0.2880 | | Acropore_tabular | 0.8900 | | Algae_assembly | 0.7410 | | Algae_drawn_up | 0.3889 | | Algae_limestone | 0.6890 | | Algae_sodding | 0.8126 | | Atra/Leucospilota | 0.6085 | | Bleached_coral | 0.6994 | | Blurred | 0.3471 | | Dead_coral | 0.6977 | | Fish | 0.6206 | | Homo_sapiens | 0.5546 | | Human_object | 0.7174 | | Living_coral | 0.6376 | | Millepore | 0.6636 | | No_acropore_encrusting | 0.5906 | | No_acropore_foliaceous | 0.7253 | | No_acropore_massive | 0.5968 | | No_acropore_solitary | 0.4364 | | No_acropore_sub_massive | 0.6084 | | Rock | 0.8513 | | Rubble | 0.7116 | | Sand | 0.8955 | | Sea_cucumber | 0.6009 | | Sea_urchins | 0.5445 | | Sponge | 0.3689 | | Syringodium_isoetifolium | 0.9401 | | Thalassodendron_ciliatum | 0.9547 | | Useless | 0.9686 | --- # Model description BeitEAU is a model built on top of BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze model for underwater multilabel image classification.The classification head is a combination of linear, ReLU, batch normalization, and dropout layers. The source code for training the model can be found in this [Git repository](https://github.com/SeatizenDOI/DinoVdeau). - **Developed by:** [lombardata](https://huggingface.co/lombardata), credits to [César Leblanc](https://huggingface.co/CesarLeblanc) and [Victor Illien](https://huggingface.co/groderg) --- # Intended uses & limitations You can use the raw model for classify diverse marine species, encompassing coral morphotypes classes taken from the Global Coral Reef Monitoring Network (GCRMN), habitats classes and seagrass species. --- # Training and evaluation data Details on the number of images for each class are given in the following table: | Class | train | test | val | Total | |:-------------------------|--------:|-------:|------:|--------:| | Acropore_branched | 1480 | 469 | 459 | 2408 | | Acropore_digitised | 571 | 156 | 161 | 888 | | Acropore_sub_massive | 150 | 52 | 41 | 243 | | Acropore_tabular | 999 | 292 | 298 | 1589 | | Algae_assembly | 2554 | 842 | 842 | 4238 | | Algae_drawn_up | 367 | 130 | 123 | 620 | | Algae_limestone | 1651 | 562 | 559 | 2772 | | Algae_sodding | 3142 | 994 | 981 | 5117 | | Atra/Leucospilota | 1084 | 349 | 359 | 1792 | | Bleached_coral | 219 | 69 | 72 | 360 | | Blurred | 191 | 68 | 61 | 320 | | Dead_coral | 1980 | 648 | 636 | 3264 | | Fish | 2018 | 661 | 642 | 3321 | | Homo_sapiens | 161 | 63 | 58 | 282 | | Human_object | 156 | 55 | 59 | 270 | | Living_coral | 397 | 151 | 153 | 701 | | Millepore | 386 | 127 | 124 | 637 | | No_acropore_encrusting | 442 | 141 | 142 | 725 | | No_acropore_foliaceous | 204 | 47 | 35 | 286 | | No_acropore_massive | 1030 | 341 | 334 | 1705 | | No_acropore_solitary | 202 | 55 | 46 | 303 | | No_acropore_sub_massive | 1402 | 428 | 426 | 2256 | | Rock | 4481 | 1495 | 1481 | 7457 | | Rubble | 3092 | 1015 | 1016 | 5123 | | Sand | 5839 | 1945 | 1935 | 9719 | | Sea_cucumber | 1407 | 437 | 450 | 2294 | | Sea_urchins | 328 | 110 | 107 | 545 | | Sponge | 267 | 98 | 105 | 470 | | Syringodium_isoetifolium | 1213 | 392 | 390 | 1995 | | Thalassodendron_ciliatum | 781 | 262 | 260 | 1303 | | Useless | 579 | 193 | 193 | 965 | --- # Training procedure ## Training hyperparameters The following hyperparameters were used during training: - **Number of Epochs**: 61.0 - **Learning Rate**: 0.001 - **Train Batch Size**: 32 - **Eval Batch Size**: 32 - **Optimizer**: Adam - **LR Scheduler Type**: ReduceLROnPlateau with a patience of 5 epochs and a factor of 0.1 - **Freeze Encoder**: Yes - **Data Augmentation**: Yes ## Data Augmentation Data were augmented using the following transformations : Train Transforms - **PreProcess**: No additional parameters - **Resize**: probability=1.00 - **RandomHorizontalFlip**: probability=0.25 - **RandomVerticalFlip**: probability=0.25 - **ColorJiggle**: probability=0.25 - **RandomPerspective**: probability=0.25 - **Normalize**: probability=1.00 Val Transforms - **PreProcess**: No additional parameters - **Resize**: probability=1.00 - **Normalize**: probability=1.00 ## Training results Epoch | Validation Loss | Accuracy | F1 Macro | F1 Micro | Learning Rate --- | --- | --- | --- | --- | --- 1 | 0.2007167786359787 | 0.1609 | 0.6701 | 0.4414 | 0.001 2 | 0.18952292203903198 | 0.1825 | 0.7090 | 0.5352 | 0.001 3 | 0.18667148053646088 | 0.1763 | 0.7166 | 0.5614 | 0.001 4 | 0.1814328134059906 | 0.1798 | 0.7215 | 0.5564 | 0.001 5 | 0.18154974281787872 | 0.1923 | 0.7243 | 0.5929 | 0.001 6 | 0.17931246757507324 | 0.1962 | 0.7348 | 0.5779 | 0.001 7 | 0.17549261450767517 | 0.1997 | 0.7318 | 0.5828 | 0.001 8 | 0.17576082050800323 | 0.2059 | 0.7272 | 0.5792 | 0.001 9 | 0.1740318238735199 | 0.1937 | 0.7299 | 0.5864 | 0.001 10 | 0.17477667331695557 | 0.1874 | 0.7276 | 0.5768 | 0.001 11 | 0.17347504198551178 | 0.1979 | 0.7381 | 0.6032 | 0.001 12 | 0.17222067713737488 | 0.2112 | 0.7353 | 0.5857 | 0.001 13 | 0.1720178872346878 | 0.2161 | 0.7369 | 0.5801 | 0.001 14 | 0.17544052004814148 | 0.1923 | 0.7266 | 0.5724 | 0.001 15 | 0.1705772429704666 | 0.1986 | 0.7413 | 0.5973 | 0.001 16 | 0.17385560274124146 | 0.2059 | 0.7282 | 0.6014 | 0.001 17 | 0.17162470519542694 | 0.1997 | 0.7463 | 0.6158 | 0.001 18 | 0.17156463861465454 | 0.2042 | 0.7337 | 0.5882 | 0.001 19 | 0.174124613404274 | 0.2094 | 0.7237 | 0.5876 | 0.001 20 | 0.1717967540025711 | 0.2094 | 0.7294 | 0.5808 | 0.001 21 | 0.1721898466348648 | 0.1986 | 0.7369 | 0.5918 | 0.001 22 | 0.16664884984493256 | 0.2115 | 0.7486 | 0.6174 | 0.0001 23 | 0.16572968661785126 | 0.2059 | 0.7474 | 0.6164 | 0.0001 24 | 0.16604110598564148 | 0.2077 | 0.7473 | 0.6207 | 0.0001 25 | 0.16543170809745789 | 0.2119 | 0.7474 | 0.6136 | 0.0001 26 | 0.16543741524219513 | 0.2098 | 0.7498 | 0.6169 | 0.0001 27 | 0.1656235158443451 | 0.2115 | 0.7497 | 0.6185 | 0.0001 28 | 0.16566258668899536 | 0.2101 | 0.7455 | 0.6091 | 0.0001 29 | 0.1653388887643814 | 0.2077 | 0.7467 | 0.6130 | 0.0001 30 | 0.16557003557682037 | 0.2119 | 0.7484 | 0.6168 | 0.0001 31 | 0.16546830534934998 | 0.2115 | 0.7498 | 0.6121 | 0.0001 32 | 0.1653573364019394 | 0.2073 | 0.7466 | 0.6116 | 0.0001 33 | 0.16520579159259796 | 0.2147 | 0.7470 | 0.6166 | 0.0001 34 | 0.1652197241783142 | 0.2188 | 0.7467 | 0.6174 | 0.0001 35 | 0.16497600078582764 | 0.2115 | 0.7503 | 0.6175 | 0.0001 36 | 0.16519583761692047 | 0.2105 | 0.7479 | 0.6171 | 0.0001 37 | 0.1649632453918457 | 0.2154 | 0.7489 | 0.6121 | 0.0001 38 | 0.16533540189266205 | 0.2164 | 0.7489 | 0.6105 | 0.0001 39 | 0.16515697538852692 | 0.2164 | 0.7506 | 0.6167 | 0.0001 40 | 0.16513320803642273 | 0.2140 | 0.7511 | 0.6164 | 0.0001 41 | 0.16498203575611115 | 0.2129 | 0.7508 | 0.6149 | 0.0001 42 | 0.16477563977241516 | 0.2136 | 0.7512 | 0.6145 | 1e-05 43 | 0.16469572484493256 | 0.2122 | 0.7490 | 0.6136 | 1e-05 44 | 0.16464821994304657 | 0.2126 | 0.7502 | 0.6148 | 1e-05 45 | 0.16469135880470276 | 0.2140 | 0.7512 | 0.6161 | 1e-05 46 | 0.16468331217765808 | 0.2126 | 0.7511 | 0.6148 | 1e-05 47 | 0.16469669342041016 | 0.2136 | 0.7510 | 0.6149 | 1e-05 48 | 0.16463501751422882 | 0.2126 | 0.7500 | 0.6143 | 1e-05 49 | 0.16467586159706116 | 0.2126 | 0.7508 | 0.6153 | 1e-05 50 | 0.16459208726882935 | 0.2126 | 0.7494 | 0.6139 | 1e-05 51 | 0.1645309180021286 | 0.2129 | 0.7498 | 0.6143 | 1e-05 52 | 0.16459544003009796 | 0.2133 | 0.7504 | 0.6142 | 1e-05 53 | 0.16463778913021088 | 0.2122 | 0.7505 | 0.6146 | 1e-05 54 | 0.16458478569984436 | 0.2129 | 0.7504 | 0.6158 | 1e-05 55 | 0.16461443901062012 | 0.2119 | 0.7505 | 0.6153 | 1e-05 56 | 0.1645844429731369 | 0.2115 | 0.7504 | 0.6145 | 1e-05 57 | 0.16459982097148895 | 0.2115 | 0.7501 | 0.6140 | 1e-05 58 | 0.16459773480892181 | 0.2115 | 0.7500 | 0.6138 | 1.0000000000000002e-06 59 | 0.164586141705513 | 0.2126 | 0.7502 | 0.6139 | 1.0000000000000002e-06 60 | 0.16457903385162354 | 0.2126 | 0.7502 | 0.6139 | 1.0000000000000002e-06 61 | 0.1645755171775818 | 0.2126 | 0.7505 | 0.6143 | 1.0000000000000002e-06 --- # Framework Versions - **Transformers**: 4.56.0.dev0 - **Pytorch**: 2.6.0+cu124 - **Datasets**: 3.0.2 - **Tokenizers**: 0.21.0