README.md

---
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