SegForCoral-2025_05_12_76513-bs16_refine is a fine-tuned version of nvidia/mit-b0.

Model description

SegForCoral-2025_05_12_76513-bs16_refine is a model built on top of nvidia/mit-b0 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.

Developed by: lombardata, credits to César Leblanc and Victor Illien

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 procedure

Training hyperparameters

The following hyperparameters were used during training:

Number of Epochs: 51.0
Learning Rate: 1e-05
Train Batch Size: 16
Eval Batch Size: 16
Optimizer: Adam
LR Scheduler Type: ReduceLROnPlateau with a patience of 5 epochs and a factor of 0.1
Freeze Encoder: Yes
Data Augmentation: Yes

Training results

Epoch	Validation Loss	Accuracy
1	N/A	1e-05
2	N/A	1e-05
3	N/A	1e-05
4	N/A	1e-05
5	N/A	1e-05
6	N/A	1e-05
7	N/A	1e-05
8	N/A	1e-05
9	N/A	1e-05
10	N/A	1e-05
11	N/A	1e-05
12	N/A	1e-05
13	N/A	1e-05
14	N/A	1e-05
15	N/A	1e-05
16	N/A	1e-05
17	N/A	1e-05
18	N/A	1e-05
19	N/A	1e-05
20	N/A	1e-05
21	N/A	1e-05
22	N/A	1e-05
23	N/A	1e-05
24	N/A	1e-05
25	N/A	1e-05
26	N/A	1e-05
27	N/A	1e-05
28	N/A	1e-05
29	N/A	1e-05
30	N/A	1e-05
31	N/A	1e-05
32	N/A	1e-05
33	N/A	1e-05
34	N/A	1e-05
35	N/A	1e-05
36	N/A	1.0000000000000002e-06
37	N/A	1.0000000000000002e-06
38	N/A	1.0000000000000002e-06
39	N/A	1.0000000000000002e-06
40	N/A	1.0000000000000002e-06
41	N/A	1.0000000000000002e-06
42	N/A	1.0000000000000002e-06
43	N/A	1.0000000000000002e-06
44	N/A	1.0000000000000002e-06
45	N/A	1.0000000000000002e-06
46	N/A	1.0000000000000002e-06
47	N/A	1.0000000000000002e-06
48	N/A	1.0000000000000002e-07
49	N/A	1.0000000000000002e-07
50	N/A	1.0000000000000002e-07
51	N/A	1.0000000000000002e-07

Framework Versions

Transformers: 4.49.0
Pytorch: 2.3.1+cu121
Datasets: 3.5.0
Tokenizers: 0.21.1