File size: 9,416 Bytes
027dee0
c2dc5e3
 
027dee0
 
c2dc5e3
 
027dee0
 
 
694bc73
027dee0
 
 
 
 
 
 
 
 
 
694bc73
027dee0
c2dc5e3
 
 
 
 
027dee0
 
 
 
694bc73
 
027dee0
 
 
694bc73
027dee0
 
 
694bc73
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
027dee0
 
 
694bc73
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
027dee0
 
 
 
 
 
 
 
694bc73
 
027dee0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
---
language:
- eng
license: apache-2.0
tags:
- multilabel-image-classification
- multilabel
- generated_from_trainer
metrics:
- accuracy
base_model: microsoft/resnet-50
model-index:
- name: resnet-50-linearhead-2024_03_12-with_data_aug_batch-size32_epochs93_freeze
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# resnet-50-linearhead-2024_03_12-with_data_aug_batch-size32_epochs93_freeze

DinoVd'eau is a fine-tuned version of [microsoft/resnet-50](https://huggingface.co/microsoft/resnet-50) on the multilabel_complete_dataset dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1518
- F1 Micro: 0.7545
- F1 Macro: 0.6309
- Roc Auc: 0.8276
- Accuracy: 0.4069
- Learning Rate: 1e-05

## 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.
- **Developed by:** [lombardata](https://huggingface.co/lombardata), credits to [César Leblanc](https://huggingface.co/CesarLeblanc)

## 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 | 804 | 202 | 200 | 1206 | 
 | Acropore_digitised | 465 | 108 | 101 | 674 | 
 | Acropore_tabular | 964 | 276 | 267 | 1507 | 
 | Algae_assembly | 2172 | 692 | 698 | 3562 | 
 | Algae_limestone | 1327 | 439 | 441 | 2207 | 
 | Algae_sodding | 2079 | 676 | 671 | 3426 | 
 | Dead_coral | 1126 | 358 | 355 | 1839 | 
 | Fish | 874 | 243 | 242 | 1359 | 
 | Human_object | 407 | 135 | 136 | 678 | 
 | Living_coral | 1765 | 580 | 571 | 2916 | 
 | Millepore | 350 | 119 | 102 | 571 | 
 | No_acropore_encrusting | 411 | 142 | 129 | 682 | 
 | No_acropore_foliaceous | 212 | 34 | 39 | 285 | 
 | No_acropore_massive | 921 | 317 | 310 | 1548 | 
 | No_acropore_sub_massive | 1205 | 362 | 363 | 1930 | 
 | Rock | 3736 | 1218 | 1217 | 6171 | 
 | Sand | 3594 | 1202 | 1194 | 5990 | 
 | Scrap | 2121 | 724 | 741 | 3586 | 
 | Sea_cucumber | 781 | 254 | 265 | 1300 | 
 | Sea_urchins | 189 | 60 | 72 | 321 | 
 | Sponge | 226 | 75 | 88 | 389 | 
 | Syringodium_isoetifolium | 1171 | 386 | 392 | 1949 | 
 | Thalassodendron_ciliatum | 783 | 261 | 260 | 1304 | 
 | Useless | 587 | 195 | 195 | 977 | 


## 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: 93

### Training results

| Training Loss | Epoch | Step  | Validation Loss | F1 Micro | F1 Macro | Roc Auc | Accuracy | Rate   |
|:-------------:|:-----:|:-----:|:---------------:|:--------:|:--------:|:-------:|:--------:|:------:|
| No log        | 1.0   | 274   | 0.2237          | 0.5839   | 0.2834   | 0.7176  | 0.1952   | 0.001  |
| 0.2683        | 2.0   | 548   | 0.1895          | 0.6773   | 0.4549   | 0.7743  | 0.3055   | 0.001  |
| 0.2683        | 3.0   | 822   | 0.1786          | 0.7021   | 0.5202   | 0.7911  | 0.3539   | 0.001  |
| 0.2058        | 4.0   | 1096  | 0.1715          | 0.7198   | 0.5666   | 0.8058  | 0.3667   | 0.001  |
| 0.2058        | 5.0   | 1370  | 0.1662          | 0.7220   | 0.5718   | 0.8050  | 0.3768   | 0.001  |
| 0.1916        | 6.0   | 1644  | 0.1648          | 0.7155   | 0.5721   | 0.7980  | 0.3796   | 0.001  |
| 0.1916        | 7.0   | 1918  | 0.1618          | 0.7281   | 0.5973   | 0.8082  | 0.3810   | 0.001  |
| 0.1858        | 8.0   | 2192  | 0.1598          | 0.7375   | 0.6061   | 0.8166  | 0.3855   | 0.001  |
| 0.1858        | 9.0   | 2466  | 0.1599          | 0.7440   | 0.6209   | 0.8223  | 0.3911   | 0.001  |
| 0.1839        | 10.0  | 2740  | 0.1584          | 0.7382   | 0.6047   | 0.8173  | 0.3949   | 0.001  |
| 0.1815        | 11.0  | 3014  | 0.1569          | 0.7414   | 0.6068   | 0.8186  | 0.3960   | 0.001  |
| 0.1815        | 12.0  | 3288  | 0.1585          | 0.7257   | 0.5953   | 0.8043  | 0.3963   | 0.001  |
| 0.1807        | 13.0  | 3562  | 0.1581          | 0.7514   | 0.6286   | 0.8311  | 0.3967   | 0.001  |
| 0.1807        | 14.0  | 3836  | 0.1565          | 0.7453   | 0.6230   | 0.8224  | 0.4022   | 0.001  |
| 0.1795        | 15.0  | 4110  | 0.1549          | 0.7504   | 0.6253   | 0.8262  | 0.3991   | 0.001  |
| 0.1795        | 16.0  | 4384  | 0.1573          | 0.7446   | 0.6290   | 0.8214  | 0.3939   | 0.001  |
| 0.178         | 17.0  | 4658  | 0.1551          | 0.7519   | 0.6287   | 0.8281  | 0.4026   | 0.001  |
| 0.178         | 18.0  | 4932  | 0.1570          | 0.7430   | 0.6155   | 0.8203  | 0.3914   | 0.001  |
| 0.1764        | 19.0  | 5206  | 0.1558          | 0.7480   | 0.6287   | 0.8236  | 0.3991   | 0.001  |
| 0.1764        | 20.0  | 5480  | 0.1574          | 0.7403   | 0.6085   | 0.8164  | 0.4001   | 0.001  |
| 0.1775        | 21.0  | 5754  | 0.1561          | 0.7532   | 0.6246   | 0.8302  | 0.4029   | 0.001  |
| 0.177         | 22.0  | 6028  | 0.1545          | 0.7596   | 0.6431   | 0.8378  | 0.3974   | 0.0001 |
| 0.177         | 23.0  | 6302  | 0.1556          | 0.7472   | 0.6292   | 0.8233  | 0.4026   | 0.0001 |
| 0.1762        | 24.0  | 6576  | 0.1548          | 0.7528   | 0.6343   | 0.8283  | 0.3994   | 0.0001 |
| 0.1762        | 25.0  | 6850  | 0.1554          | 0.7468   | 0.6225   | 0.8222  | 0.3994   | 0.0001 |
| 0.1759        | 26.0  | 7124  | 0.1548          | 0.7529   | 0.6326   | 0.8297  | 0.3977   | 0.0001 |
| 0.1759        | 27.0  | 7398  | 0.1552          | 0.7516   | 0.6352   | 0.8282  | 0.3970   | 0.0001 |
| 0.1752        | 28.0  | 7672  | 0.1543          | 0.7523   | 0.6328   | 0.8277  | 0.4092   | 0.0001 |
| 0.1752        | 29.0  | 7946  | 0.1545          | 0.7506   | 0.6312   | 0.8265  | 0.4019   | 0.0001 |
| 0.1757        | 30.0  | 8220  | 0.1550          | 0.7554   | 0.6394   | 0.8340  | 0.4040   | 0.0001 |
| 0.1757        | 31.0  | 8494  | 0.1554          | 0.7512   | 0.6345   | 0.8279  | 0.4022   | 0.0001 |
| 0.1758        | 32.0  | 8768  | 0.1545          | 0.7513   | 0.6302   | 0.8275  | 0.4033   | 0.0001 |
| 0.1755        | 33.0  | 9042  | 0.1555          | 0.7456   | 0.6261   | 0.8223  | 0.3977   | 0.0001 |
| 0.1755        | 34.0  | 9316  | 0.1533          | 0.7515   | 0.6307   | 0.8260  | 0.4109   | 0.0001 |
| 0.1752        | 35.0  | 9590  | 0.1551          | 0.7506   | 0.6325   | 0.8261  | 0.4054   | 0.0001 |
| 0.1752        | 36.0  | 9864  | 0.1530          | 0.7539   | 0.6299   | 0.8287  | 0.4026   | 0.0001 |
| 0.1752        | 37.0  | 10138 | 0.1546          | 0.7464   | 0.6270   | 0.8223  | 0.4036   | 0.0001 |
| 0.1752        | 38.0  | 10412 | 0.1549          | 0.7539   | 0.6364   | 0.8314  | 0.3987   | 0.0001 |
| 0.1763        | 39.0  | 10686 | 0.1547          | 0.7579   | 0.6421   | 0.8361  | 0.3977   | 0.0001 |
| 0.1763        | 40.0  | 10960 | 0.1544          | 0.7539   | 0.6345   | 0.8302  | 0.4005   | 0.0001 |
| 0.176         | 41.0  | 11234 | 0.1557          | 0.7536   | 0.6347   | 0.8298  | 0.4015   | 0.0001 |
| 0.1758        | 42.0  | 11508 | 0.1540          | 0.7474   | 0.6277   | 0.8226  | 0.3960   | 0.0001 |
| 0.1758        | 43.0  | 11782 | 0.1548          | 0.7578   | 0.6384   | 0.8374  | 0.3970   | 1e-05  |
| 0.1764        | 44.0  | 12056 | 0.1543          | 0.7582   | 0.6398   | 0.8352  | 0.4012   | 1e-05  |
| 0.1764        | 45.0  | 12330 | 0.1544          | 0.7448   | 0.6206   | 0.8196  | 0.3991   | 1e-05  |
| 0.1746        | 46.0  | 12604 | 0.1546          | 0.7452   | 0.6223   | 0.8208  | 0.4050   | 1e-05  |


### Framework versions

- Transformers 4.36.2
- Pytorch 2.1.0+cu118
- Datasets 2.14.5
- Tokenizers 0.15.0