You can deploy this model and start earning money today!

groderg/
ClipEAU-vit-base-patch32-2025_11_03_55901-bs32_freeze

Name: ClipEAU-vit-base-patch32-2025_11_03_55901-bs32_freeze
Availability: InStock

tensorboard

safetensors

clip

multilabel-image-classification

multilabel

generated_from_trainer

eng

cc0-1.0

Views

No views yet

ClipEAU is a fine-tuned version of ClipEAU-vit-base-patch32-2025_11_03_55901-bs32_freeze. It achieves the following results on the test set:

Loss: 0.1981
F1 Micro: 0.6811
F1 Macro: 0.4890
Accuracy: 0.1745

Class	F1 per class
Acropore_branched	0.6435
Acropore_digitised	0.3536
Acropore_sub_massive	0.1806
Acropore_tabular	0.7672
Algae_assembly	0.7188
Algae_drawn_up	0.3009
Algae_limestone	0.6345
Algae_sodding	0.7843
Atra/Leucospilota	0.4893
Bleached_coral	0.5588
Blurred	0.1871
Dead_coral	0.6294
Fish	0.5666
Homo_sapiens	0.5225
Human_object	0.6444
Living_coral	0.5072
Millepore	0.4047
No_acropore_encrusting	0.5487
No_acropore_foliaceous	0.6034
No_acropore_massive	0.5000
No_acropore_solitary	0.2667
No_acropore_sub_massive	0.5328
Rock	0.8336
Rubble	0.6817
Sand	0.8700
Sea_cucumber	0.5231
Sea_urchins	0.4375
Sponge	0.3228
Syringodium_isoetifolium	0.8362
Thalassodendron_ciliatum	0.7737
Useless	0.8626

Model description

ClipEAU is a model built on top of ClipEAU-vit-base-patch32-2025_11_03_55901-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: 89.0
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	0.2755436599254608	0.0548	0.4870	0.1841	0.001
2	0.27313947677612305	0.1061	0.4816	0.2190	0.001
3	0.24732808768749237	0.1092	0.5586	0.2896	0.001
4	0.24704696238040924	0.1131	0.5864	0.3132	0.001
5	0.23763391375541687	0.1155	0.5996	0.3203	0.001
6	0.24488729238510132	0.1005	0.5455	0.2758	0.001
7	0.23577998578548431	0.1316	0.5706	0.3076	0.001
8	0.23298335075378418	0.1257	0.5880	0.3409	0.001
9	0.2437519133090973	0.1201	0.5783	0.3233	0.001
10	0.2337779998779297	0.1257	0.5844	0.3481	0.001
11	0.22841598093509674	0.1281	0.5768	0.3405	0.001
12	0.23676042258739471	0.1295	0.6101	0.3479	0.001
13	0.22685056924819946	0.1323	0.6174	0.3855	0.001
14	0.22243528068065643	0.1354	0.6163	0.3304	0.001
15	0.2308577597141266	0.1309	0.6026	0.3426	0.001
16	0.22836312651634216	0.1379	0.5836	0.3525	0.001
17	0.22632907330989838	0.1211	0.6431	0.4095	0.001
18	0.22169864177703857	0.1431	0.6216	0.4108	0.001
19	0.22429880499839783	0.1469	0.6158	0.3903	0.001
20	0.22168485820293427	0.1497	0.6244	0.4109	0.001
21	0.2181965708732605	0.1585	0.6275	0.4087	0.001
22	0.21742750704288483	0.1459	0.6402	0.4088	0.001
23	0.21984589099884033	0.1508	0.6372	0.4019	0.001
24	0.21336357295513153	0.1431	0.6349	0.4086	0.001
25	0.215973362326622	0.1581	0.6435	0.4073	0.001
26	0.21738594770431519	0.1627	0.6428	0.4412	0.001
27	0.21718686819076538	0.1515	0.6318	0.4376	0.001
28	0.2167457938194275	0.1469	0.6548	0.4502	0.001
29	0.21596720814704895	0.1602	0.6239	0.3974	0.001
30	0.21519586443901062	0.1609	0.6316	0.4272	0.001
31	0.20102238655090332	0.1742	0.6709	0.4708	0.0001
32	0.20028482377529144	0.1728	0.6652	0.4765	0.0001
33	0.2016177624464035	0.1658	0.6587	0.4668	0.0001
34	0.20070569217205048	0.1703	0.6587	0.4671	0.0001
35	0.19858147203922272	0.1749	0.6693	0.4791	0.0001
36	0.19878427684307098	0.1696	0.6671	0.4713	0.0001
37	0.19879288971424103	0.1728	0.6705	0.4824	0.0001
38	0.2002326399087906	0.1756	0.6662	0.4798	0.0001
39	0.19776451587677002	0.1759	0.6704	0.4807	0.0001
40	0.1989012360572815	0.1735	0.6743	0.4826	0.0001
41	0.19757546484470367	0.1735	0.6745	0.4863	0.0001
42	0.19867026805877686	0.1742	0.6701	0.4850	0.0001
43	0.1974734216928482	0.1794	0.6762	0.4756	0.0001
44	0.19798807799816132	0.1805	0.6740	0.4820	0.0001
45	0.19826450943946838	0.1784	0.6741	0.4701	0.0001
46	0.19794772565364838	0.1787	0.6769	0.4901	0.0001
47	0.19794227182865143	0.1787	0.6791	0.4937	0.0001
48	0.19741104543209076	0.1759	0.6829	0.4981	0.0001
49	0.20017848908901215	0.1798	0.6696	0.4878	0.0001
50	0.19828735291957855	0.1773	0.6832	0.4995	0.0001
51	0.19745203852653503	0.1846	0.6785	0.4869	0.0001
52	0.19813166558742523	0.1728	0.6823	0.4954	0.0001
53	0.19777563214302063	0.1808	0.6811	0.5010	0.0001
54	0.19714485108852386	0.1777	0.6781	0.4998	0.0001
55	0.19739548861980438	0.1745	0.6773	0.5015	0.0001
56	0.19639085233211517	0.1784	0.6791	0.4859	0.0001
57	0.19614720344543457	0.1770	0.6806	0.4934	0.0001
58	0.19727133214473724	0.1763	0.6767	0.4840	0.0001
59	0.19701360166072845	0.1846	0.6769	0.4857	0.0001
60	0.19576512277126312	0.1777	0.6786	0.4898	0.0001
61	0.19700495898723602	0.1805	0.6852	0.5093	0.0001
62	0.19687828421592712	0.1766	0.6834	0.5009	0.0001
63	0.1967148333787918	0.1752	0.6849	0.4928	0.0001
64	0.19667862355709076	0.1805	0.6781	0.4918	0.0001
65	0.196905255317688	0.1766	0.6804	0.4911	0.0001
66	0.19522123038768768	0.1770	0.6867	0.5141	0.0001
67	0.196974515914917	0.1780	0.6882	0.5127	0.0001
68	0.19643187522888184	0.1791	0.6864	0.4956	0.0001
69	0.19658857583999634	0.1801	0.6844	0.4981	0.0001
70	0.19634096324443817	0.1794	0.6814	0.5045	0.0001
71	0.19620731472969055	0.1818	0.6840	0.4957	0.0001
72	0.19560062885284424	0.1791	0.6827	0.4968	0.0001
73	0.19446241855621338	0.1881	0.6876	0.5035	1e-05
74	0.19424258172512054	0.1850	0.6874	0.5037	1e-05
75	0.19434040784835815	0.1839	0.6882	0.5094	1e-05
76	0.19450798630714417	0.1850	0.6868	0.5042	1e-05
77	0.19437134265899658	0.1885	0.6882	0.5141	1e-05
78	0.1945873498916626	0.1871	0.6889	0.5078	1e-05
79	0.19400206208229065	0.1885	0.6889	0.5101	1e-05
80	0.19423776865005493	0.1867	0.6893	0.5123	1e-05
81	0.1944921761751175	0.1878	0.6877	0.5087	1e-05
82	0.19430173933506012	0.1846	0.6872	0.5094	1e-05
83	0.19456203281879425	0.1832	0.6894	0.5132	1e-05
84	0.19433410465717316	0.1850	0.6899	0.5093	1e-05
85	0.1942111849784851	0.1846	0.6872	0.5088	1e-05
86	0.19422636926174164	0.1857	0.6889	0.5119	1.0000000000000002e-06
87	0.19421224296092987	0.1839	0.6895	0.5124	1.0000000000000002e-06
88	0.19428525865077972	0.1867	0.6892	0.5130	1.0000000000000002e-06
89	0.19422417879104614	0.1853	0.6894	0.5122	1.0000000000000002e-06

Framework Versions

Transformers: 4.56.0.dev0
Pytorch: 2.6.0+cu124
Datasets: 3.0.2
Tokenizers: 0.21.0

groderg/ClipEAU-vit-base-patch32-2025_11_03_55901-bs32_freeze