Hugging Face's logo

lombardata
/
drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs

TensorBoard
Safetensors
English
dinov2
multilabel-image-classification
multilabel
Generated from Trainer
Model card Files Metrics Community
lombardata commited on
Commit
e57a489
•
1 Parent(s): 5decf45

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +153 -86
README.md CHANGED
@@ -1,107 +1,174 @@
 
1
  ---
2
- license: apache-2.0
3
- base_model: facebook/dinov2-large
 
4
  tags:
 
 
5
  - generated_from_trainer
 
6
  model-index:
7
  - name: drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs
8
  results: []
9
  ---
10
 
11
- <!-- This model card has been generated automatically according to the information the Trainer had access to. You
12
- should probably proofread and complete it, then remove this comment. -->
13
 
14
- # drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs
15
 
16
- This model is a fine-tuned version of [facebook/dinov2-large](https://huggingface.co/facebook/dinov2-large) on the None dataset.
17
- It achieves the following results on the evaluation set:
18
  - Loss: 0.3261
19
- - Rmse: 0.2445
20
- - Mae: 0.1621
21
- - R2: 0.4058
22
- - Explained Variance: 0.4072
23
- - Learning Rate: 1e-05
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
- ## Model description
 
26
 
27
- More information needed
28
 
29
- ## Intended uses & limitations
30
 
31
- More information needed
 
 
 
 
 
32
 
33
- ## Training and evaluation data
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34
 
35
- More information needed
36
 
37
- ## Training procedure
38
 
39
- ### Training hyperparameters
40
 
41
  The following hyperparameters were used during training:
42
- - learning_rate: 0.001
43
- - train_batch_size: 16
44
- - eval_batch_size: 16
45
- - seed: 42
46
- - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
- - lr_scheduler_type: linear
48
- - num_epochs: 150
49
- - mixed_precision_training: Native AMP
50
-
51
- ### Training results
52
-
53
- | Training Loss | Epoch | Step | Validation Loss | Rmse | Mae | R2 | Explained Variance | Rate |
54
- |:-------------:|:-----:|:-----:|:---------------:|:------:|:------:|:------:|:------------------:|:------:|
55
- | 0.4549 | 1.0 | 721 | 0.3625 | 0.2669 | 0.1880 | 0.2744 | 0.2809 | 0.001 |
56
- | 0.3806 | 2.0 | 1442 | 0.3457 | 0.2560 | 0.1685 | 0.3367 | 0.3395 | 0.001 |
57
- | 0.368 | 3.0 | 2163 | 0.3518 | 0.2597 | 0.1747 | 0.3157 | 0.3180 | 0.001 |
58
- | 0.3637 | 4.0 | 2884 | 0.3508 | 0.2563 | 0.1751 | 0.3345 | 0.3375 | 0.001 |
59
- | 0.36 | 5.0 | 3605 | 0.3436 | 0.2546 | 0.1696 | 0.3371 | 0.3404 | 0.001 |
60
- | 0.3585 | 6.0 | 4326 | 0.3510 | 0.2598 | 0.1767 | 0.3175 | 0.3211 | 0.001 |
61
- | 0.3581 | 7.0 | 5047 | 0.3412 | 0.2538 | 0.1750 | 0.3471 | 0.3510 | 0.001 |
62
- | 0.3601 | 8.0 | 5768 | 0.3456 | 0.2561 | 0.1678 | 0.3435 | 0.3552 | 0.001 |
63
- | 0.3619 | 9.0 | 6489 | 0.3425 | 0.2545 | 0.1741 | 0.3409 | 0.3427 | 0.001 |
64
- | 0.355 | 10.0 | 7210 | 0.3396 | 0.2525 | 0.1711 | 0.3583 | 0.3602 | 0.001 |
65
- | 0.3574 | 11.0 | 7931 | 0.3448 | 0.2542 | 0.1721 | 0.3498 | 0.3524 | 0.001 |
66
- | 0.3549 | 12.0 | 8652 | 0.3416 | 0.2527 | 0.1767 | 0.3577 | 0.3604 | 0.001 |
67
- | 0.354 | 13.0 | 9373 | 0.3399 | 0.2527 | 0.1677 | 0.3523 | 0.3545 | 0.001 |
68
- | 0.3566 | 14.0 | 10094 | 0.3452 | 0.2540 | 0.1746 | 0.3443 | 0.3479 | 0.001 |
69
- | 0.3553 | 15.0 | 10815 | 0.3485 | 0.2568 | 0.1801 | 0.3333 | 0.3463 | 0.001 |
70
- | 0.3536 | 16.0 | 11536 | 0.3435 | 0.2537 | 0.1718 | 0.3473 | 0.3499 | 0.001 |
71
- | 0.3518 | 17.0 | 12257 | 0.3412 | 0.2508 | 0.1711 | 0.3633 | 0.3668 | 0.0001 |
72
- | 0.3475 | 18.0 | 12978 | 0.3399 | 0.2507 | 0.1708 | 0.3649 | 0.3656 | 0.0001 |
73
- | 0.347 | 19.0 | 13699 | 0.3333 | 0.2483 | 0.1675 | 0.3775 | 0.3787 | 0.0001 |
74
- | 0.3445 | 20.0 | 14420 | 0.3332 | 0.2478 | 0.1688 | 0.3810 | 0.3822 | 0.0001 |
75
- | 0.3447 | 21.0 | 15141 | 0.3324 | 0.2476 | 0.1673 | 0.3810 | 0.3833 | 0.0001 |
76
- | 0.3445 | 22.0 | 15862 | 0.3320 | 0.2472 | 0.1671 | 0.3836 | 0.3849 | 0.0001 |
77
- | 0.3398 | 23.0 | 16583 | 0.3301 | 0.2461 | 0.1658 | 0.3890 | 0.3900 | 0.0001 |
78
- | 0.3417 | 24.0 | 17304 | 0.3299 | 0.2458 | 0.1648 | 0.3899 | 0.3905 | 0.0001 |
79
- | 0.3406 | 25.0 | 18025 | 0.3296 | 0.2458 | 0.1641 | 0.3903 | 0.3910 | 0.0001 |
80
- | 0.3381 | 26.0 | 18746 | 0.3289 | 0.2454 | 0.1632 | 0.3926 | 0.3930 | 0.0001 |
81
- | 0.3399 | 27.0 | 19467 | 0.3304 | 0.2461 | 0.1674 | 0.3891 | 0.3908 | 0.0001 |
82
- | 0.3377 | 28.0 | 20188 | 0.3288 | 0.2451 | 0.1645 | 0.3955 | 0.3972 | 0.0001 |
83
- | 0.3384 | 29.0 | 20909 | 0.3294 | 0.2451 | 0.1656 | 0.3961 | 0.3973 | 0.0001 |
84
- | 0.3372 | 30.0 | 21630 | 0.3314 | 0.2464 | 0.1684 | 0.3914 | 0.3955 | 0.0001 |
85
- | 0.3375 | 31.0 | 22351 | 0.3291 | 0.2457 | 0.1608 | 0.3904 | 0.3936 | 0.0001 |
86
- | 0.3373 | 32.0 | 23072 | 0.3289 | 0.2453 | 0.1631 | 0.3959 | 0.3971 | 0.0001 |
87
- | 0.3362 | 33.0 | 23793 | 0.3272 | 0.2444 | 0.1628 | 0.3972 | 0.3989 | 0.0001 |
88
- | 0.3371 | 34.0 | 24514 | 0.3270 | 0.2443 | 0.1621 | 0.3976 | 0.3981 | 0.0001 |
89
- | 0.3342 | 35.0 | 25235 | 0.3264 | 0.2439 | 0.1615 | 0.3987 | 0.3997 | 0.0001 |
90
- | 0.3367 | 36.0 | 25956 | 0.3293 | 0.2455 | 0.1656 | 0.3946 | 0.3955 | 0.0001 |
91
- | 0.3363 | 37.0 | 26677 | 0.3271 | 0.2442 | 0.1597 | 0.3996 | 0.4032 | 0.0001 |
92
- | 0.3357 | 38.0 | 27398 | 0.3270 | 0.2437 | 0.1613 | 0.4022 | 0.4041 | 0.0001 |
93
- | 0.3377 | 39.0 | 28119 | 0.3326 | 0.2438 | 0.1575 | 0.4007 | 0.4027 | 0.0001 |
94
- | 0.3354 | 40.0 | 28840 | 0.3328 | 0.2442 | 0.1651 | 0.4003 | 0.4018 | 0.0001 |
95
- | 0.3363 | 41.0 | 29561 | 0.3307 | 0.2435 | 0.1627 | 0.4031 | 0.4045 | 0.0001 |
96
- | 0.335 | 42.0 | 30282 | 0.3310 | 0.2436 | 0.1641 | 0.4030 | 0.4040 | 1e-05 |
97
- | 0.334 | 43.0 | 31003 | 0.3296 | 0.2429 | 0.1603 | 0.4052 | 0.4059 | 1e-05 |
98
- | 0.3366 | 44.0 | 31724 | 0.3302 | 0.2432 | 0.1625 | 0.4038 | 0.4055 | 1e-05 |
99
- | 0.3326 | 45.0 | 32445 | 0.3266 | 0.2430 | 0.1617 | 0.4047 | 0.4055 | 1e-05 |
100
-
101
-
102
- ### Framework versions
103
-
104
- - Transformers 4.41.0
105
- - Pytorch 2.5.0+cu124
106
- - Datasets 3.0.2
107
- - Tokenizers 0.19.1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
  ---
3
+ language:
4
+ - eng
5
+ license: cc0-1.0
6
  tags:
7
+ - multilabel-image-classification
8
+ - multilabel
9
  - generated_from_trainer
10
+ base_model: drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs
11
  model-index:
12
  - name: drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs
13
  results: []
14
  ---
15
 
16
+ drone-DinoVdeau-produttoria-probabilities is a fine-tuned version of [drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs](https://huggingface.co/drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs). It achieves the following results on the test set:
 
17
 
 
18
 
 
 
19
  - Loss: 0.3261
20
+ - F1 Micro: 0.8621
21
+ - F1 Macro: 0.8264
22
+ - Accuracy: 0.1682
23
+ - RMSE: 0.2445
24
+ - MAE: 0.1621
25
+ - R2: 0.4057
26
+
27
+ | Class | F1 per class |
28
+ |----------|-------|
29
+ | Acropore_branched | 0.8063 |
30
+ | Acropore_digitised | 0.7335 |
31
+ | Acropore_tabular | 0.6247 |
32
+ | Algae | 0.9859 |
33
+ | Dead_coral | 0.8424 |
34
+ | Fish | 0.7464 |
35
+ | Millepore | 0.6453 |
36
+ | No_acropore_encrusting | 0.7292 |
37
+ | No_acropore_massive | 0.8681 |
38
+ | No_acropore_sub_massive | 0.8092 |
39
+ | Rock | 0.9925 |
40
+ | Rubble | 0.9693 |
41
+ | Sand | 0.9904 |
42
+
43
+
44
+ ---
45
 
46
+ # Model description
47
+ drone-DinoVdeau-produttoria-probabilities is a model built on top of drone-DinoVdeau-produttoria-probabilities-large-2024_11_06-batch-size16_freeze_probs model for underwater multilabel image classification.The classification head is a combination of linear, ReLU, batch normalization, and dropout layers.
48
 
49
+ The source code for training the model can be found in this [Git repository](https://github.com/SeatizenDOI/DinoVdeau).
50
 
51
+ - **Developed by:** [lombardata](https://huggingface.co/lombardata), credits to [César Leblanc](https://huggingface.co/CesarLeblanc) and [Victor Illien](https://huggingface.co/groderg)
52
 
53
+ ---
54
+
55
+ # Intended uses & limitations
56
+ 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.
57
+
58
+ ---
59
 
60
+ # Training and evaluation data
61
+ Details on the estimated number of images for each class are given in the following table:
62
+ | Class | train | test | val | Total |
63
+ |:------------------------|--------:|-------:|------:|--------:|
64
+ | Acropore_branched | 2028 | 684 | 686 | 3398 |
65
+ | Acropore_digitised | 2006 | 735 | 717 | 3458 |
66
+ | Acropore_tabular | 1237 | 461 | 451 | 2149 |
67
+ | Algae | 11086 | 3671 | 3675 | 18432 |
68
+ | Dead_coral | 6354 | 2161 | 2147 | 10662 |
69
+ | Fish | 4032 | 1430 | 1430 | 6892 |
70
+ | Millepore | 1943 | 783 | 772 | 3498 |
71
+ | No_acropore_encrusting | 2663 | 986 | 957 | 4606 |
72
+ | No_acropore_massive | 6897 | 2375 | 2375 | 11647 |
73
+ | No_acropore_sub_massive | 5416 | 1988 | 1958 | 9362 |
74
+ | Rock | 11164 | 3726 | 3725 | 18615 |
75
+ | Rubble | 10687 | 3570 | 3572 | 17829 |
76
+ | Sand | 11151 | 3726 | 3723 | 18600 |
77
 
78
+ ---
79
 
80
+ # Training procedure
81
 
82
+ ## Training hyperparameters
83
 
84
  The following hyperparameters were used during training:
85
+
86
+ - **Number of Epochs**: 45.0
87
+ - **Learning Rate**: 0.001
88
+ - **Train Batch Size**: 16
89
+ - **Eval Batch Size**: 16
90
+ - **Optimizer**: Adam
91
+ - **LR Scheduler Type**: ReduceLROnPlateau with a patience of 5 epochs and a factor of 0.1
92
+ - **Freeze Encoder**: Yes
93
+ - **Data Augmentation**: Yes
94
+
95
+
96
+ ## Data Augmentation
97
+ Data were augmented using the following transformations :
98
+
99
+ Train Transforms
100
+ - **PreProcess**: No additional parameters
101
+ - **Resize**: probability=1.00
102
+ - **RandomHorizontalFlip**: probability=0.25
103
+ - **RandomVerticalFlip**: probability=0.25
104
+ - **ColorJiggle**: probability=0.25
105
+ - **RandomPerspective**: probability=0.25
106
+ - **Normalize**: probability=1.00
107
+
108
+ Val Transforms
109
+ - **PreProcess**: No additional parameters
110
+ - **Resize**: probability=1.00
111
+ - **Normalize**: probability=1.00
112
+
113
+
114
+
115
+ ## Training results
116
+ Epoch | Validation Loss | MAE | RMSE | R2 | Learning Rate
117
+ --- | --- | --- | --- | --- | ---
118
+ 0 | N/A | 0.0000 | 0.0000 | 0.0000 | 0.001
119
+ 1 | 0.36246591806411743 | 0.1880 | 0.2669 | 0.2744 | 0.001
120
+ 2 | 0.3457428216934204 | 0.1685 | 0.2560 | 0.3367 | 0.001
121
+ 3 | 0.3518487811088562 | 0.1747 | 0.2597 | 0.3157 | 0.001
122
+ 4 | 0.3507988750934601 | 0.1751 | 0.2563 | 0.3345 | 0.001
123
+ 5 | 0.3436409533023834 | 0.1696 | 0.2546 | 0.3371 | 0.001
124
+ 6 | 0.35096481442451477 | 0.1767 | 0.2598 | 0.3175 | 0.001
125
+ 7 | 0.3412320613861084 | 0.1750 | 0.2538 | 0.3471 | 0.001
126
+ 8 | 0.3456409275531769 | 0.1678 | 0.2561 | 0.3435 | 0.001
127
+ 9 | 0.3425351679325104 | 0.1741 | 0.2545 | 0.3409 | 0.001
128
+ 10 | 0.33964109420776367 | 0.1711 | 0.2525 | 0.3583 | 0.001
129
+ 11 | 0.34479108452796936 | 0.1721 | 0.2542 | 0.3498 | 0.001
130
+ 12 | 0.3415849804878235 | 0.1767 | 0.2527 | 0.3577 | 0.001
131
+ 13 | 0.33990854024887085 | 0.1677 | 0.2527 | 0.3523 | 0.001
132
+ 14 | 0.34520208835601807 | 0.1746 | 0.2540 | 0.3443 | 0.001
133
+ 15 | 0.34849879145622253 | 0.1801 | 0.2568 | 0.3333 | 0.001
134
+ 16 | 0.34347954392433167 | 0.1718 | 0.2537 | 0.3473 | 0.001
135
+ 17 | 0.341246634721756 | 0.1711 | 0.2508 | 0.3633 | 0.0001
136
+ 18 | 0.3398562967777252 | 0.1708 | 0.2507 | 0.3649 | 0.0001
137
+ 19 | 0.3332718312740326 | 0.1675 | 0.2483 | 0.3775 | 0.0001
138
+ 20 | 0.333162784576416 | 0.1688 | 0.2478 | 0.3810 | 0.0001
139
+ 21 | 0.3324449062347412 | 0.1673 | 0.2476 | 0.3810 | 0.0001
140
+ 22 | 0.3320053517818451 | 0.1671 | 0.2472 | 0.3836 | 0.0001
141
+ 23 | 0.3301050662994385 | 0.1658 | 0.2461 | 0.3890 | 0.0001
142
+ 24 | 0.3298528492450714 | 0.1648 | 0.2458 | 0.3899 | 0.0001
143
+ 25 | 0.32962867617607117 | 0.1641 | 0.2458 | 0.3903 | 0.0001
144
+ 26 | 0.32889437675476074 | 0.1632 | 0.2454 | 0.3926 | 0.0001
145
+ 27 | 0.33042922616004944 | 0.1674 | 0.2461 | 0.3891 | 0.0001
146
+ 28 | 0.32880541682243347 | 0.1645 | 0.2451 | 0.3955 | 0.0001
147
+ 29 | 0.3293789327144623 | 0.1656 | 0.2451 | 0.3961 | 0.0001
148
+ 30 | 0.33135533332824707 | 0.1684 | 0.2464 | 0.3914 | 0.0001
149
+ 31 | 0.32911789417266846 | 0.1608 | 0.2457 | 0.3904 | 0.0001
150
+ 32 | 0.3289436399936676 | 0.1631 | 0.2453 | 0.3959 | 0.0001
151
+ 33 | 0.3271527588367462 | 0.1628 | 0.2444 | 0.3972 | 0.0001
152
+ 34 | 0.32699429988861084 | 0.1621 | 0.2443 | 0.3976 | 0.0001
153
+ 35 | 0.32638314366340637 | 0.1615 | 0.2439 | 0.3987 | 0.0001
154
+ 36 | 0.3293066918849945 | 0.1656 | 0.2455 | 0.3946 | 0.0001
155
+ 37 | 0.3271186649799347 | 0.1597 | 0.2442 | 0.3996 | 0.0001
156
+ 38 | 0.32695677876472473 | 0.1613 | 0.2437 | 0.4022 | 0.0001
157
+ 39 | 0.33263665437698364 | 0.1575 | 0.2438 | 0.4007 | 0.0001
158
+ 40 | 0.33278176188468933 | 0.1651 | 0.2442 | 0.4003 | 0.0001
159
+ 41 | 0.33069443702697754 | 0.1627 | 0.2435 | 0.4031 | 0.0001
160
+ 42 | 0.3310275375843048 | 0.1641 | 0.2436 | 0.4030 | 1e-05
161
+ 43 | 0.32956016063690186 | 0.1603 | 0.2429 | 0.4052 | 1e-05
162
+ 44 | 0.33022987842559814 | 0.1625 | 0.2432 | 0.4038 | 1e-05
163
+ 45 | 0.3266430199146271 | 0.1617 | 0.2430 | 0.4047 | 1e-05
164
+
165
+
166
+ ---
167
+
168
+ # Framework Versions
169
+
170
+ - **Transformers**: 4.41.0
171
+ - **Pytorch**: 2.5.0+cu124
172
+ - **Datasets**: 3.0.2
173
+ - **Tokenizers**: 0.19.1
174
+