swin-tiny-patch4-window7-224-finetuned-eurosat

This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.8360
• Accuracy: 0.7664

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.9333	7	3.8894	0.0841
3.897	2.0	15	3.8185	0.0841
3.8553	2.9333	22	3.7402	0.0748
3.7568	4.0	30	3.6372	0.0748
3.7568	4.9333	37	3.5482	0.0841
3.5912	6.0	45	3.4069	0.1121
3.4342	6.9333	52	3.2939	0.1308
3.2601	8.0	60	3.1786	0.2150
3.2601	8.9333	67	3.0323	0.2336
3.0498	10.0	75	2.8695	0.2617
2.849	10.9333	82	2.8505	0.2523
2.6452	12.0	90	2.6319	0.2804
2.6452	12.9333	97	2.4654	0.3271
2.4123	14.0	105	2.3995	0.3364
2.2561	14.9333	112	2.2584	0.4019
2.0447	16.0	120	2.2000	0.4299
2.0447	16.9333	127	2.0806	0.4393
1.8569	18.0	135	2.0593	0.4393
1.7447	18.9333	142	1.8832	0.4673
1.5821	20.0	150	1.8218	0.5047
1.5821	20.9333	157	1.7334	0.5421
1.3999	22.0	165	1.6213	0.5514
1.2901	22.9333	172	1.5932	0.5234
1.1569	24.0	180	1.5256	0.5701
1.1569	24.9333	187	1.4281	0.5888
1.0903	26.0	195	1.3997	0.5794
0.9674	26.9333	202	1.4017	0.5888
0.98	28.0	210	1.2916	0.5981
0.98	28.9333	217	1.3018	0.5981
0.8772	30.0	225	1.2552	0.6355
0.7842	30.9333	232	1.2372	0.6075
0.7438	32.0	240	1.1908	0.6168
0.7438	32.9333	247	1.1567	0.6636
0.725	34.0	255	1.1542	0.6262
0.6709	34.9333	262	1.1377	0.6262
0.6898	36.0	270	1.0524	0.6636
0.6898	36.9333	277	1.0272	0.6729
0.6125	38.0	285	1.0399	0.6355
0.6153	38.9333	292	1.0308	0.6822
0.5898	40.0	300	1.0151	0.7009
0.5898	40.9333	307	1.0483	0.6542
0.5881	42.0	315	0.9926	0.7009
0.54	42.9333	322	1.0300	0.6916
0.4515	44.0	330	0.9262	0.7383
0.4515	44.9333	337	0.9486	0.7290
0.5057	46.0	345	0.9219	0.7103
0.4905	46.9333	352	1.0184	0.6822
0.4669	48.0	360	0.9337	0.7290
0.4669	48.9333	367	0.9431	0.7103
0.4437	50.0	375	0.9312	0.7009
0.4754	50.9333	382	0.9245	0.7196
0.4119	52.0	390	0.8826	0.7383
0.4119	52.9333	397	0.9262	0.7196
0.4087	54.0	405	0.8882	0.7477
0.3987	54.9333	412	0.9282	0.7290
0.4253	56.0	420	0.9004	0.7477
0.4253	56.9333	427	0.8783	0.7477
0.4134	58.0	435	0.8360	0.7664
0.4024	58.9333	442	0.9016	0.7196
0.3688	60.0	450	0.9251	0.6822
0.3688	60.9333	457	0.9086	0.7103
0.3833	62.0	465	0.8494	0.7383
0.3614	62.9333	472	0.8299	0.7290
0.3792	64.0	480	0.9015	0.7383
0.3792	64.9333	487	0.8802	0.7196
0.3632	66.0	495	0.8881	0.7009
0.3405	66.9333	502	0.8578	0.7383
0.3673	68.0	510	0.8540	0.7570
0.3673	68.9333	517	0.8345	0.7383
0.3379	70.0	525	0.7919	0.7383
0.3389	70.9333	532	0.8384	0.7290
0.3363	72.0	540	0.8306	0.7383
0.3363	72.9333	547	0.8875	0.7477
0.3494	74.0	555	0.9151	0.7009
0.2989	74.9333	562	0.8606	0.7103
0.3157	76.0	570	0.8640	0.7383
0.3157	76.9333	577	0.8532	0.7290
0.3013	78.0	585	0.8479	0.7103
0.2968	78.9333	592	0.8839	0.7383
0.3013	80.0	600	0.8837	0.7196
0.3013	80.9333	607	0.8694	0.7103
0.3247	82.0	615	0.8721	0.7290
0.2515	82.9333	622	0.8605	0.7290
0.3175	84.0	630	0.8505	0.7290
0.3175	84.9333	637	0.8488	0.7290
0.3015	86.0	645	0.8554	0.7383
0.2989	86.9333	652	0.8707	0.7290
0.3155	88.0	660	0.8712	0.7290
0.3155	88.9333	667	0.8659	0.7290
0.2871	90.0	675	0.8573	0.7290
0.2872	90.9333	682	0.8530	0.7290
0.2587	92.0	690	0.8516	0.7383
0.2587	92.9333	697	0.8502	0.7383
0.3133	93.3333	700	0.8501	0.7383

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1