Hugging Face's logo

lombardata
/
DinoVdeau-large-2024_04_03-with_data_aug_batch-size32_epochs150_freeze

Transformers
TensorBoard
Safetensors
English
dinov2
multilabel-image-classification
multilabel
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
DinoVdeau-large-2024_04_03-with_data_aug_batch-size32_epochs150_freeze / README.md
lombardata's picture
lombardata
Update README.md
41c2bc0 verified
preview code
|
raw
history blame
18.2 kB
metadata
language:
  - eng
license: apache-2.0
tags:
  - multilabel-image-classification
  - multilabel
  - generated_from_trainer
metrics:
  - accuracy
base_model: facebook/dinov2-large
model-index:
  - name: DinoVdeau-large-2024_04_03-with_data_aug_batch-size32_epochs150_freeze
    results: []

DinoVdeau-large-2024_04_03-with_data_aug_batch-size32_epochs150_freeze

DinoVd'eau is a fine-tuned version of facebook/dinov2-large on the multilabel_complete_dataset dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1181
  • F1 Micro: 0.8219
  • F1 Macro: 0.7131
  • Roc Auc: 0.8797
  • Accuracy: 0.3214
  • Learning Rate: 0.0000

Model description

DinoVd'eau is a model built on top of dinov2 model for underwater multilabel image classification.The classification head is a combination of linear, ReLU, batch normalization, and dropout layers.

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:

train val test Total
Acropore_branched 1504 445 430 2379
Acropore_digitised 593 151 144 888
Acropore_sub_massive 148 54 41 243
Acropore_tabular 1012 290 287 1589
Algae_assembly 2545 858 835 4238
Algae_drawn_up 376 123 121 620
Algae_limestone 1652 561 559 2772
Algae_sodding 3094 1011 1012 5117
Atra/Leucospilota 1081 352 359 1792
Bleached_coral 220 70 70 360
Blurred 192 62 66 320
Dead_coral 2001 637 626 3264
Fish 2068 611 642 3321
Homo_sapiens 162 60 60 282
Human_object 157 60 53 270
Living_coral 147 56 47 250
Millepore 378 131 128 637
No_acropore_encrusting 422 152 151 725
No_acropore_foliaceous 200 46 40 286
No_acropore_massive 1033 337 335 1705
No_acropore_solitary 193 56 54 303
No_acropore_sub_massive 1412 418 426 2256
Rock 4487 1481 1489 7457
Sand 5806 1959 1954 9719
Scrap 3063 1030 1030 5123
Sea_cucumber 1396 453 445 2294
Sea_urchins 319 122 104 545
Sponge 273 107 90 470
Syringodium_isoetifolium 1198 399 398 1995
Thalassodendron_ciliatum 781 260 262 1303
Useless 579 193 193 965

Training procedure

Data Augmentation

Data were augmented using the following transformations :

  • training transformations : Sequential( (0): PreProcess() (1): Resize(output_size=(518, 518), p=1.0, p_batch=1.0, same_on_batch=True, size=(518, 518), side=short, resample=bilinear, align_corners=True, antialias=False) (2): RandomHorizontalFlip(p=0.25, p_batch=1.0, same_on_batch=False) (3): RandomVerticalFlip(p=0.25, p_batch=1.0, same_on_batch=False) (4): ColorJiggle(brightness=0.0, contrast=0.0, saturation=0.0, hue=0.0, p=0.25, p_batch=1.0, same_on_batch=False) (5): RandomPerspective(distortion_scale=0.5, p=0.25, p_batch=1.0, same_on_batch=False, align_corners=False, resample=bilinear) (6): Normalize(p=1.0, p_batch=1.0, same_on_batch=True, mean=tensor([0.4850, 0.4560, 0.4060]), std=tensor([0.2290, 0.2240, 0.2250])) )
  • validation transformations : Sequential( (0): PreProcess() (1): Resize(output_size=(518, 518), p=1.0, p_batch=1.0, same_on_batch=True, size=(518, 518), side=short, resample=bilinear, align_corners=True, antialias=False) (2): Normalize(p=1.0, p_batch=1.0, same_on_batch=True, mean=tensor([0.4850, 0.4560, 0.4060]), std=tensor([0.2290, 0.2240, 0.2250])) )

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: ReduceLROnPlateau with a patience of 5 epochs and a factor of 0.1
  • freeze_encoder: True
  • num_epochs: 150

Training results

Training Loss Epoch Step Accuracy F1 Macro F1 Micro Validation Loss Roc Auc Rate
No log 1.0 271 0.2207 0.4907 0.7369 0.1679 0.8188 0.001
0.2713 2.0 542 0.2515 0.5389 0.7614 0.1540 0.8356 0.001
0.2713 3.0 813 0.2526 0.6054 0.7728 0.1477 0.8472 0.001
0.1679 4.0 1084 0.2594 0.5848 0.7755 0.1578 0.8442 0.001
0.1679 5.0 1355 0.2618 0.6125 0.7819 0.1426 0.8555 0.001
0.1598 6.0 1626 0.2550 0.6239 0.7822 0.1422 0.8542 0.001
0.1598 7.0 1897 0.2557 0.6320 0.7825 0.1426 0.8534 0.001
0.1571 8.0 2168 0.2629 0.6223 0.7756 0.1528 0.8437 0.001
0.1571 9.0 2439 0.2481 0.6413 0.7796 0.1438 0.8549 0.001
0.1554 10.0 2710 0.2697 0.6289 0.7889 0.1405 0.8621 0.001
0.1554 11.0 2981 0.2684 0.6222 0.7898 0.1409 0.8614 0.001
0.1536 12.0 3252 0.2725 0.6166 0.7863 0.1392 0.8528 0.001
0.1526 13.0 3523 0.2625 0.6419 0.7877 0.1399 0.8559 0.001
0.1526 14.0 3794 0.2649 0.6326 0.7860 0.1438 0.8609 0.001
0.1535 15.0 4065 0.2735 0.6499 0.7930 0.1377 0.8625 0.001
0.1535 16.0 4336 0.2677 0.6435 0.7868 0.1397 0.8526 0.001
0.1517 17.0 4607 0.2646 0.6401 0.7928 0.1382 0.8634 0.001
0.1517 18.0 4878 0.2684 0.6286 0.7912 0.1392 0.8624 0.001
0.1524 19.0 5149 0.2636 0.6183 0.7874 0.1392 0.8576 0.001
0.1524 20.0 5420 0.2598 0.6286 0.7878 0.1386 0.8578 0.001
0.1527 21.0 5691 0.2601 0.6408 0.7880 0.1374 0.8557 0.001
0.1527 22.0 5962 0.2704 0.6476 0.7897 0.1377 0.8577 0.001
0.1513 23.0 6233 0.2697 0.6443 0.7955 0.1373 0.8655 0.001
0.1514 24.0 6504 0.2656 0.6477 0.7877 0.1593 0.8547 0.001
0.1514 25.0 6775 0.2656 0.6477 0.7909 0.1371 0.8619 0.001
0.1513 26.0 7046 0.2666 0.6273 0.7871 0.1374 0.8535 0.001
0.1513 27.0 7317 0.2646 0.6470 0.7934 0.1373 0.8595 0.001
0.1508 28.0 7588 0.2735 0.6523 0.7933 0.1353 0.8584 0.001
0.1508 29.0 7859 0.2776 0.6522 0.7960 0.1362 0.8645 0.001
0.1506 30.0 8130 0.2505 0.6283 0.7849 0.1384 0.8546 0.001
0.1506 31.0 8401 0.2718 0.6630 0.7964 0.1342 0.8636 0.001
0.151 32.0 8672 0.2718 0.6556 0.7968 0.1366 0.8696 0.001
0.151 33.0 8943 0.2824 0.6635 0.7985 0.1359 0.8701 0.001
0.1507 34.0 9214 0.2814 0.6400 0.7999 0.1335 0.8657 0.001
0.1507 35.0 9485 0.2725 0.6520 0.7963 0.1343 0.8653 0.001
0.1495 36.0 9756 0.2636 0.6451 0.7924 0.1429 0.8626 0.001
0.1496 37.0 10027 0.2732 0.6531 0.7981 0.1331 0.8638 0.001
0.1496 38.0 10298 0.2684 0.6306 0.7938 0.1350 0.8617 0.001
0.1503 39.0 10569 0.2800 0.6465 0.7984 0.1352 0.8661 0.001
0.1503 40.0 10840 0.2728 0.6271 0.7925 0.1347 0.8594 0.001
0.1505 41.0 11111 0.2721 0.6601 0.7935 0.1340 0.8579 0.001
0.1505 42.0 11382 0.2711 0.6636 0.7983 0.1322 0.8652 0.001
0.1491 43.0 11653 0.2735 0.6493 0.7949 0.1360 0.8635 0.001
0.1491 44.0 11924 0.2814 0.6400 0.7955 0.1361 0.8625 0.001
0.1507 45.0 12195 0.2814 0.6424 0.7971 0.1328 0.8640 0.001
0.1507 46.0 12466 0.2787 0.6469 0.7939 0.1328 0.8581 0.001
0.1495 47.0 12737 0.2752 0.6351 0.7977 0.1332 0.8672 0.001
0.1498 48.0 13008 0.2817 0.6490 0.8013 0.1325 0.8694 0.001
0.1498 49.0 13279 0.2883 0.6738 0.8062 0.1283 0.8710 0.0001
0.1416 50.0 13550 0.2872 0.6734 0.8087 0.1287 0.8747 0.0001
0.1416 51.0 13821 0.2900 0.6714 0.8067 0.1280 0.8706 0.0001
0.1387 52.0 14092 0.2900 0.6744 0.8067 0.1262 0.8702 0.0001
0.1387 53.0 14363 0.2910 0.6764 0.8094 0.1262 0.8729 0.0001
0.1356 54.0 14634 0.2948 0.6744 0.8091 0.1257 0.8702 0.0001
0.1356 55.0 14905 0.1257 0.8106 0.6814 0.8742 0.2948 0.0001
0.1348 56.0 15176 0.1260 0.8108 0.6772 0.8738 0.3010 0.0001
0.1348 57.0 15447 0.1250 0.8129 0.6806 0.8768 0.2986 0.0001
0.135 58.0 15718 0.1242 0.8142 0.6859 0.8762 0.3082 0.0001
0.135 59.0 15989 0.1245 0.8124 0.6870 0.8763 0.3027 0.0001
0.1334 60.0 16260 0.1242 0.8138 0.6854 0.8772 0.3030 0.0001
0.1335 61.0 16531 0.1240 0.8140 0.6889 0.8756 0.3065 0.0001
0.1335 62.0 16802 0.1249 0.8152 0.6809 0.8798 0.3016 0.0001
0.1308 63.0 17073 0.1233 0.8146 0.6848 0.8757 0.3068 0.0001
0.1308 64.0 17344 0.1234 0.8151 0.6908 0.8769 0.3058 0.0001
0.1326 65.0 17615 0.1233 0.8124 0.6812 0.8735 0.3034 0.0001
0.1326 66.0 17886 0.1232 0.8145 0.6878 0.8788 0.3027 0.0001
0.1306 67.0 18157 0.1228 0.8115 0.6857 0.8707 0.3075 0.0001
0.1306 68.0 18428 0.1226 0.8153 0.6913 0.8767 0.3075 0.0001
0.1299 69.0 18699 0.1227 0.8143 0.6764 0.8751 0.3085 0.0001
0.1299 70.0 18970 0.1230 0.8187 0.6999 0.8838 0.3106 0.0001
0.1295 71.0 19241 0.1225 0.8153 0.6893 0.8756 0.3068 0.0001
0.1289 72.0 19512 0.1223 0.8151 0.6868 0.8776 0.3037 0.0001
0.1289 73.0 19783 0.1223 0.8165 0.6918 0.8782 0.3054 0.0001
0.1279 74.0 20054 0.1225 0.8143 0.6856 0.8747 0.3054 0.0001
0.1279 75.0 20325 0.1221 0.8167 0.6878 0.8784 0.3102 0.0001
0.1276 76.0 20596 0.1217 0.8190 0.6964 0.8812 0.3167 0.0001
0.1276 77.0 20867 0.1217 0.8179 0.6940 0.8796 0.3102 0.0001
0.1274 78.0 21138 0.1216 0.8143 0.6859 0.8735 0.3082 0.0001
0.1274 79.0 21409 0.1215 0.8165 0.6945 0.8766 0.3147 0.0001
0.1269 80.0 21680 0.1214 0.8193 0.6999 0.8803 0.3147 0.0001
0.1269 81.0 21951 0.1214 0.8194 0.6974 0.8828 0.3113 0.0001
0.1259 82.0 22222 0.1212 0.8171 0.6956 0.8782 0.3102 0.0001
0.1259 83.0 22493 0.1208 0.8190 0.6970 0.8791 0.3123 0.0001
0.1258 84.0 22764 0.1209 0.8204 0.6997 0.8813 0.3154 0.0001
0.1251 85.0 23035 0.1211 0.8163 0.6935 0.8752 0.3065 0.0001
0.1251 86.0 23306 0.1203 0.8201 0.6972 0.8804 0.3154 0.0001
0.1251 87.0 23577 0.1208 0.8182 0.6947 0.8785 0.3150 0.0001
0.1251 88.0 23848 0.1214 0.8181 0.6937 0.8788 0.3154 0.0001
0.1246 89.0 24119 0.1206 0.8201 0.6953 0.8797 0.3106 0.0001
0.1246 90.0 24390 0.1210 0.8214 0.6960 0.8819 0.3164 0.0001
0.1239 91.0 24661 0.1199 0.8202 0.7006 0.8805 0.3154 0.0001
0.1239 92.0 24932 0.1208 0.8222 0.7039 0.8856 0.3161 0.0001
0.1238 93.0 25203 0.1204 0.8199 0.7004 0.8808 0.3133 0.0001
0.1238 94.0 25474 0.1200 0.8230 0.7036 0.8847 0.3143 0.0001
0.1237 95.0 25745 0.1206 0.8209 0.7069 0.8817 0.3188 0.0001
0.1234 96.0 26016 0.1201 0.8222 0.7060 0.8820 0.3147 0.0001
0.1234 97.0 26287 0.1204 0.8208 0.7074 0.8830 0.3092 0.0001
0.1215 98.0 26558 0.1200 0.8241 0.7125 0.8859 0.3188 1e-05
0.1215 99.0 26829 0.1195 0.8247 0.7127 0.8864 0.3171 1e-05
0.1208 100.0 27100 0.1192 0.8225 0.7077 0.8818 0.3164 1e-05
0.1208 101.0 27371 0.1193 0.8232 0.7060 0.8831 0.3171 1e-05
0.1195 102.0 27642 0.1197 0.8238 0.7105 0.8848 0.3185 1e-05
0.1195 103.0 27913 0.1191 0.8216 0.7076 0.8805 0.3140 1e-05
0.1197 104.0 28184 0.1193 0.8239 0.7063 0.8843 0.3202 1e-05
0.1197 105.0 28455 0.1190 0.8213 0.7071 0.8799 0.3126 1e-05
0.1189 106.0 28726 0.1190 0.8233 0.7061 0.8835 0.3202 1e-05
0.1189 107.0 28997 0.1194 0.8224 0.7038 0.8811 0.3164 1e-05
0.1194 108.0 29268 0.1191 0.8232 0.7110 0.8830 0.3191 1e-05
0.1187 109.0 29539 0.1189 0.8230 0.7101 0.8817 0.3174 1e-05
0.1187 110.0 29810 0.1192 0.8224 0.7044 0.8810 0.3161 1e-05
0.1185 111.0 30081 0.1192 0.8226 0.7083 0.8827 0.3174 1e-05
0.1185 112.0 30352 0.1190 0.8239 0.7093 0.8841 0.3205 1e-05
0.119 113.0 30623 0.1195 0.8233 0.7080 0.8845 0.3171 1e-05
0.119 114.0 30894 0.1190 0.8220 0.7062 0.8799 0.3181 1e-05
0.1182 115.0 31165 0.1192 0.8229 0.7081 0.8823 0.3174 1e-05
0.1182 116.0 31436 0.1190 0.8256 0.7128 0.8862 0.3250 0.0000
0.1191 117.0 31707 0.1187 0.8231 0.7104 0.8821 0.3171 0.0000
0.1191 118.0 31978 0.1189 0.8236 0.7061 0.8830 0.3198 0.0000
0.1179 119.0 32249 0.1189 0.8233 0.7080 0.8830 0.3181 0.0000
0.1176 120.0 32520 0.1190 0.8239 0.7101 0.8838 0.3185 0.0000
0.1176 121.0 32791 0.1195 0.8254 0.7128 0.8872 0.3209 0.0000
0.1175 122.0 33062 0.1192 0.8223 0.7048 0.8813 0.3154 0.0000
0.1175 123.0 33333 0.1192 0.8255 0.7154 0.8856 0.3212 0.0000
0.1176 124.0 33604 0.1189 0.8239 0.7109 0.8837 0.3209 0.0000
0.1176 125.0 33875 0.1189 0.8252 0.7102 0.8847 0.3226 0.0000
0.1179 126.0 34146 0.1189 0.8206 0.7025 0.8787 0.3164 0.0000
0.1179 127.0 34417 0.1190 0.8245 0.7104 0.8839 0.3216 0.0000

Framework versions

  • Transformers 4.36.2
  • Pytorch 2.1.0+cu118
  • Datasets 2.18.0
  • Tokenizers 0.15.0