You can deploy this model and start earning money today!

groderg/
BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze

Name: BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze
Availability: InStock

tensorboard

safetensors

beit

multilabel-image-classification

multilabel

generated_from_trainer

eng

cc0-1.0

Views

No views yet

BeitEAU is a fine-tuned version of 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.

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 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

groderg/BeitEAU-base-patch16-384-2025_11_07_78282-bs32_freeze