File size: 8,461 Bytes
528b0b1
2ae096e
528b0b1
 
 
2ae096e
528b0b1
 
2ae096e
528b0b1
2ae096e
 
 
 
 
528b0b1
2ae096e
 
 
528b0b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2ae096e
528b0b1
 
2ae096e
528b0b1
2ae096e
528b0b1
2ae096e
528b0b1
 
 
 
 
 
2ae096e
528b0b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2ae096e
528b0b1
2ae096e
528b0b1
2ae096e
528b0b1
2ae096e
 
528b0b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198

---
language:
- eng
license: cc0-1.0
tags:
- multilabel-image-classification
- multilabel
- generated_from_trainer
base_model: drone-DinoVdeau-produttoria-probabilities-large-2024_11_04-batch-size64_freeze_probs
model-index:
- name: drone-DinoVdeau-produttoria-probabilities-large-2024_11_04-batch-size64_freeze_probs
  results: []
---

drone-DinoVdeau-produttoria-probabilities is a fine-tuned version of [drone-DinoVdeau-produttoria-probabilities-large-2024_11_04-batch-size64_freeze_probs](https://huggingface.co/drone-DinoVdeau-produttoria-probabilities-large-2024_11_04-batch-size64_freeze_probs). It achieves the following results on the test set:


- Loss: 0.3194
- F1 Micro: 0.8663
- F1 Macro: 0.8311
- Accuracy: 0.1799
- RMSE: 0.2404
- MAE: 0.1536
- R2: 0.4282

| Class | F1 per class |
|----------|-------|
| Acropore_branched | 0.8010 |
| Acropore_digitised | 0.7454 |
| Acropore_tabular | 0.6426 |
| Algae | 0.9852 |
| Dead_coral | 0.8448 |
| Fish | 0.7497 |
| Millepore | 0.6641 |
| No_acropore_encrusting | 0.7391 |
| No_acropore_massive | 0.8688 |
| No_acropore_sub_massive | 0.8137 |
| Rock | 0.9924 |
| Rubble | 0.9691 |
| Sand | 0.9888 |


---

# Model description
drone-DinoVdeau-produttoria-probabilities is a model built on top of drone-DinoVdeau-produttoria-probabilities-large-2024_11_04-batch-size64_freeze_probs 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 estimated number of images for each class are given in the following table:
| Class                   |   train |   test |   val |   Total |
|:------------------------|--------:|-------:|------:|--------:|
| Acropore_branched       |    2028 |    684 |   686 |    3398 |
| Acropore_digitised      |    2006 |    735 |   717 |    3458 |
| Acropore_tabular        |    1237 |    461 |   451 |    2149 |
| Algae                   |   11086 |   3671 |  3675 |   18432 |
| Dead_coral              |    6354 |   2161 |  2147 |   10662 |
| Fish                    |    4032 |   1430 |  1430 |    6892 |
| Millepore               |    1943 |    783 |   772 |    3498 |
| No_acropore_encrusting  |    2663 |    986 |   957 |    4606 |
| No_acropore_massive     |    6897 |   2375 |  2375 |   11647 |
| No_acropore_sub_massive |    5416 |   1988 |  1958 |    9362 |
| Rock                    |   11164 |   3726 |  3725 |   18615 |
| Rubble                  |   10687 |   3570 |  3572 |   17829 |
| Sand                    |   11151 |   3726 |  3723 |   18600 |

---

# Training procedure

## Training hyperparameters

The following hyperparameters were used during training:

- **Number of Epochs**: 69.0
- **Learning Rate**: 0.001
- **Train Batch Size**: 64
- **Eval Batch Size**: 64
- **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 | MAE | RMSE | R2 | Learning Rate
--- | --- | --- | --- | --- | ---
1 | 0.36101797223091125 | 0.1878 | 0.2645 | 0.2818 | 0.001
2 | 0.3464529514312744 | 0.1778 | 0.2566 | 0.3349 | 0.001
3 | 0.34133487939834595 | 0.1731 | 0.2532 | 0.3536 | 0.001
4 | 0.3406243324279785 | 0.1743 | 0.2532 | 0.3519 | 0.001
5 | 0.3341675400733948 | 0.1661 | 0.2496 | 0.3702 | 0.001
6 | 0.33848145604133606 | 0.1739 | 0.2512 | 0.3651 | 0.001
7 | 0.3320676386356354 | 0.1650 | 0.2476 | 0.3836 | 0.001
8 | 0.3332081437110901 | 0.1629 | 0.2484 | 0.3802 | 0.001
9 | 0.3305376172065735 | 0.1652 | 0.2468 | 0.3859 | 0.001
10 | 0.33136793971061707 | 0.1655 | 0.2476 | 0.3827 | 0.001
11 | 0.3319685757160187 | 0.1602 | 0.2474 | 0.3840 | 0.001
12 | 0.3341500759124756 | 0.1683 | 0.2494 | 0.3761 | 0.001
13 | 0.33248215913772583 | 0.1649 | 0.2480 | 0.3821 | 0.001
14 | 0.33228376507759094 | 0.1700 | 0.2472 | 0.3878 | 0.001
15 | 0.334873229265213 | 0.1703 | 0.2493 | 0.3749 | 0.001
16 | 0.3279329538345337 | 0.1649 | 0.2448 | 0.3983 | 0.0001
17 | 0.3279244005680084 | 0.1648 | 0.2448 | 0.3984 | 0.0001
18 | 0.3262367248535156 | 0.1622 | 0.2440 | 0.4025 | 0.0001
19 | 0.3247373402118683 | 0.1588 | 0.2432 | 0.4046 | 0.0001
20 | 0.32612329721450806 | 0.1625 | 0.2433 | 0.4059 | 0.0001
21 | 0.3241129517555237 | 0.1606 | 0.2424 | 0.4095 | 0.0001
22 | 0.32355180382728577 | 0.1587 | 0.2422 | 0.4111 | 0.0001
23 | 0.3242079019546509 | 0.1601 | 0.2423 | 0.4107 | 0.0001
24 | 0.3227241337299347 | 0.1586 | 0.2414 | 0.4150 | 0.0001
25 | 0.3223778307437897 | 0.1587 | 0.2413 | 0.4148 | 0.0001
26 | 0.3217927813529968 | 0.1557 | 0.2413 | 0.4143 | 0.0001
27 | 0.3227355182170868 | 0.1603 | 0.2416 | 0.4138 | 0.0001
28 | 0.32067713141441345 | 0.1562 | 0.2405 | 0.4186 | 0.0001
29 | 0.32205939292907715 | 0.1597 | 0.2411 | 0.4163 | 0.0001
30 | 0.32246074080467224 | 0.1608 | 0.2413 | 0.4164 | 0.0001
31 | 0.3223503530025482 | 0.1535 | 0.2416 | 0.4134 | 0.0001
32 | 0.3212696313858032 | 0.1553 | 0.2408 | 0.4180 | 0.0001
33 | 0.32156360149383545 | 0.1583 | 0.2414 | 0.4123 | 0.0001
34 | 0.3205103278160095 | 0.1562 | 0.2406 | 0.4172 | 0.0001
35 | 0.3197581171989441 | 0.1535 | 0.2399 | 0.4215 | 0.0001
36 | 0.3211075961589813 | 0.1577 | 0.2406 | 0.4187 | 0.0001
37 | 0.3203599154949188 | 0.1520 | 0.2403 | 0.4188 | 0.0001
38 | 0.32143038511276245 | 0.1560 | 0.2409 | 0.4170 | 0.0001
39 | 0.3195198178291321 | 0.1520 | 0.2397 | 0.4226 | 0.0001
40 | 0.3207896649837494 | 0.1577 | 0.2404 | 0.4204 | 0.0001
41 | 0.3197501003742218 | 0.1547 | 0.2398 | 0.4217 | 0.0001
42 | 0.32175716757774353 | 0.1589 | 0.2410 | 0.4174 | 0.0001
43 | 0.3189575970172882 | 0.1544 | 0.2396 | 0.4235 | 0.0001
44 | 0.31898385286331177 | 0.1534 | 0.2396 | 0.4230 | 0.0001
45 | 0.31977778673171997 | 0.1566 | 0.2397 | 0.4239 | 0.0001
46 | 0.3193351626396179 | 0.1556 | 0.2398 | 0.4213 | 0.0001
47 | 0.31895366311073303 | 0.1524 | 0.2393 | 0.4245 | 0.0001
48 | 0.3192996680736542 | 0.1525 | 0.2398 | 0.4215 | 0.0001
49 | 0.32073548436164856 | 0.1558 | 0.2405 | 0.4187 | 0.0001
50 | 0.3198453485965729 | 0.1572 | 0.2400 | 0.4218 | 1e-05
51 | 0.32436585426330566 | 0.1602 | 0.2426 | 0.4092 | 1e-05
52 | 0.31899821758270264 | 0.1550 | 0.2396 | 0.4227 | 1e-05
53 | 0.31892043352127075 | 0.1552 | 0.2394 | 0.4249 | 1e-05
54 | 0.3194037675857544 | 0.1540 | 0.2396 | 0.4227 | 1e-05
55 | 0.3184601366519928 | 0.1539 | 0.2391 | 0.4250 | 1e-05
56 | 0.318115234375 | 0.1527 | 0.2388 | 0.4273 | 1e-05
57 | 0.31871330738067627 | 0.1532 | 0.2392 | 0.4259 | 1e-05
58 | 0.32010164856910706 | 0.1567 | 0.2401 | 0.4217 | 1e-05
59 | 0.31807705760002136 | 0.1528 | 0.2388 | 0.4270 | 1e-05
60 | 0.3181913495063782 | 0.1534 | 0.2389 | 0.4256 | 1e-05
61 | 0.3185857832431793 | 0.1510 | 0.2391 | 0.4255 | 1e-05
62 | 0.32031872868537903 | 0.1596 | 0.2398 | 0.4240 | 1e-05
63 | 0.31964218616485596 | 0.1570 | 0.2397 | 0.4242 | 1e-05
64 | 0.31808170676231384 | 0.1527 | 0.2391 | 0.4244 | 1e-05
65 | 0.31850185990333557 | 0.1550 | 0.2390 | 0.4259 | 1e-05
66 | 0.3186076879501343 | 0.1562 | 0.2389 | 0.4278 | 1.0000000000000002e-06
67 | 0.3181016743183136 | 0.1526 | 0.2387 | 0.4270 | 1.0000000000000002e-06
68 | 0.3194774389266968 | 0.1549 | 0.2397 | 0.4221 | 1.0000000000000002e-06
69 | 0.3183264136314392 | 0.1540 | 0.2390 | 0.4259 | 1.0000000000000002e-06


---

# Framework Versions

- **Transformers**: 4.41.0
- **Pytorch**: 2.5.0+cu124
- **Datasets**: 3.0.2
- **Tokenizers**: 0.19.1