SegFormer B5-Finetuned on Larger Images: An Open-Source Image Segmentation Model - Accurately Distinguish Background and Branch Regions in Large-Size Images

Segformerb5 Finetuned Largerimages

Developed by JCAI2000

An image segmentation model based on the SegFormer-B5 architecture, fine-tuned on larger-sized image datasets, excels at distinguishing between background and branch regions

Image Segmentation

Transformers

Open Source License:Other #High-precision image segmentation #Remote sensing image processing #SegFormer architecture

Downloads 14

Release Time : 9/25/2023

Model Overview

This model is a fine-tuned version of SegFormer-B5, specifically optimized for larger-sized images, primarily used for image segmentation tasks, with particular expertise in distinguishing background and branch regions.

Model Features

High-precision background segmentation

Achieves 99.1% background accuracy and 98.24% background IoU on the evaluation set

Optimized branch detection

Specifically optimized for branch regions, achieving 72.69% branch accuracy

Large-size image processing

Fine-tuned specifically on larger-sized image datasets

Model Capabilities

Image segmentation

Background recognition

Branch region detection

Use Cases

Medical imaging

Vascular branch analysis

Used for identifying and segmenting vascular branches in medical images

Can accurately distinguish vascular branches from background tissues

Remote sensing

Road network extraction

Extracting road networks from satellite or aerial images

Can effectively identify road branch structures

🚀 segformerb5-finetuned-largerImages

This model is a fine - tuned version of the base model, which can effectively perform image segmentation tasks on the JCAI2000/LargerImagesLabelled dataset, achieving high accuracy and performance.

🚀 Quick Start

This model is a fine - tuned version of [JCAI2000/segformer - b5 - finetuned - 100by100PNG - 50epochs - attempt2 - 100epochs - backgroundclass - 2](https://huggingface.co/JCAI2000/segformer - b5 - finetuned - 100by100PNG - 50epochs - attempt2 - 100epochs - backgroundclass - 2) on the JCAI2000/LargerImagesLabelled dataset. It achieves the following results on the evaluation set:

Loss: 0.0724
Mean Iou: 0.7754
Mean Accuracy: 0.8589
Overall Accuracy: 0.9828
Accuracy Background: 0.9910
Accuracy Branch: 0.7269
Iou Background: 0.9824
Iou Branch: 0.5684

📚 Documentation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e - 05
train_batch_size: 2
eval_batch_size: 2
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Branch	Iou Background	Iou Branch
0.0872	1.18	20	0.0678	0.6870	0.7241	0.9784	0.9954	0.4528	0.9781	0.3958
0.0848	2.35	40	0.0577	0.7333	0.7908	0.9806	0.9932	0.5884	0.9802	0.4864
0.0549	3.53	60	0.0634	0.7485	0.8857	0.9773	0.9834	0.7879	0.9767	0.5203
0.0653	4.71	80	0.0493	0.7662	0.8346	0.9827	0.9926	0.6767	0.9823	0.5500
0.0596	5.88	100	0.0476	0.7497	0.7920	0.9828	0.9955	0.5885	0.9825	0.5168
0.0458	7.06	120	0.0478	0.7636	0.8357	0.9823	0.9921	0.6793	0.9819	0.5452
0.0285	8.24	140	0.0458	0.7758	0.8574	0.9829	0.9913	0.7235	0.9825	0.5691
0.0341	9.41	160	0.0466	0.7670	0.8376	0.9827	0.9923	0.6829	0.9823	0.5517
0.0369	10.59	180	0.0491	0.7699	0.8731	0.9813	0.9885	0.7576	0.9809	0.5589
0.0352	11.76	200	0.0465	0.7731	0.8551	0.9826	0.9911	0.7191	0.9822	0.5640
0.0477	12.94	220	0.0462	0.7721	0.8415	0.9832	0.9926	0.6905	0.9828	0.5615
0.0404	14.12	240	0.0493	0.7704	0.8734	0.9814	0.9886	0.7583	0.9809	0.5599
0.0221	15.29	260	0.0458	0.7798	0.8719	0.9828	0.9901	0.7536	0.9823	0.5772
0.0263	16.47	280	0.0450	0.7778	0.8509	0.9835	0.9923	0.7096	0.9831	0.5726
0.0364	17.65	300	0.0489	0.7756	0.8537	0.9830	0.9917	0.7158	0.9827	0.5686
0.02	18.82	320	0.0493	0.7713	0.8474	0.9828	0.9918	0.7031	0.9824	0.5602
0.0193	20.0	340	0.0481	0.7786	0.8694	0.9827	0.9903	0.7484	0.9823	0.5749
0.0133	21.18	360	0.0486	0.7756	0.8552	0.9830	0.9915	0.7189	0.9826	0.5686
0.0163	22.35	380	0.0492	0.7768	0.8632	0.9828	0.9907	0.7357	0.9824	0.5713
0.0252	23.53	400	0.0510	0.7725	0.8605	0.9823	0.9904	0.7306	0.9819	0.5632
0.0178	24.71	420	0.0509	0.7770	0.8665	0.9826	0.9904	0.7427	0.9822	0.5719
0.0167	25.88	440	0.0516	0.7748	0.8615	0.9826	0.9906	0.7323	0.9822	0.5675
0.0332	27.06	460	0.0507	0.7702	0.8422	0.9829	0.9922	0.6921	0.9825	0.5578
0.021	28.24	480	0.0522	0.7710	0.8482	0.9827	0.9916	0.7048	0.9823	0.5597
0.0284	29.41	500	0.0536	0.7762	0.8643	0.9826	0.9905	0.7380	0.9822	0.5702
0.0174	30.59	520	0.0535	0.7739	0.8591	0.9826	0.9908	0.7274	0.9822	0.5657
0.0228	31.76	540	0.0527	0.7765	0.8578	0.9830	0.9913	0.7243	0.9826	0.5703
0.0347	32.94	560	0.0530	0.7754	0.8534	0.9830	0.9917	0.7151	0.9826	0.5681
0.0182	34.12	580	0.0554	0.7764	0.8670	0.9825	0.9902	0.7437	0.9821	0.5706
0.0139	35.29	600	0.0521	0.7786	0.8540	0.9834	0.9920	0.7159	0.9830	0.5742
0.0121	36.47	620	0.0549	0.7787	0.8728	0.9826	0.9899	0.7557	0.9822	0.5752
0.0191	37.65	640	0.0560	0.7766	0.8678	0.9825	0.9902	0.7454	0.9821	0.5711
0.0216	38.82	660	0.0553	0.7745	0.8543	0.9829	0.9914	0.7172	0.9825	0.5665
0.0135	40.0	680	0.0569	0.7738	0.8640	0.9823	0.9902	0.7379	0.9819	0.5658
0.0167	41.18	700	0.0566	0.7765	0.8619	0.9828	0.9908	0.7330	0.9824	0.5707
0.0224	42.35	720	0.0570	0.7768	0.8680	0.9825	0.9902	0.7458	0.9821	0.5714
0.0188	43.53	740	0.0575	0.7768	0.8630	0.9828	0.9907	0.7353	0.9824	0.5713
0.0338	44.71	760	0.0565	0.7783	0.8634	0.9830	0.9909	0.7359	0.9826	0.5741
0.0122	45.88	780	0.0585	0.7788	0.8656	0.9829	0.9907	0.7404	0.9825	0.5750
0.0119	47.06	800	0.0587	0.7774	0.8639	0.9828	0.9907	0.7371	0.9824	0.5725
0.0086	48.24	820	0.0594	0.7777	0.8567	0.9832	0.9916	0.7218	0.9828	0.5726
0.0094	49.41	840	0.0597	0.7766	0.8627	0.9828	0.9907	0.7347	0.9823	0.5708
0.0107	50.59	860	0.0619	0.7773	0.8624	0.9829	0.9909	0.7338	0.9825	0.5722
0.0175	51.76	880	0.0605	0.7752	0.8588	0.9828	0.9910	0.7266	0.9824	0.5681
0.0139	52.94	900	0.0620	0.7786	0.8675	0.9828	0.9905	0.7446	0.9824	0.5748
0.02	54.12	920	0.0633	0.7764	0.8628	0.9827	0.9907	0.7348	0.9823	0.5704
0.0294	55.29	940	0.0637	0.7762	0.8584	0.9829	0.9912	0.7256	0.9825	0.5699
0.0175	56.47	960	0.0639	0.7789	0.8717	0.9827	0.9900	0.7534	0.9822	0.5755
0.008	57.65	980	0.0640	0.7797	0.8667	0.9830	0.9907	0.7428	0.9826	0.5768
0.0132	58.82	1000	0.0652	0.7754	0.8657	0.9825	0.9902	0.7412	0.9820	0.5687
0.0339	60.0	1020	0.0640	0.7785	0.8664	0.9828	0.9906	0.7421	0.9824	0.5746
0.0144	61.18	1040	0.0618	0.7804	0.8577	0.9835	0.9919	0.7236	0.9831	0.5778
0.0206	62.35	1060	0.0653	0.7767	0.8636	0.9827	0.9907	0.7366	0.9823	0.5710
0.0165	63.53	1080	0.0651	0.7774	0.8602	0.9830	0.9912	0.7293	0.9826	0.5722
0.0175	64.71	1100	0.0648	0.7758	0.8568	0.9829	0.9913	0.7222	0.9825	0.5690
0.0104	65.88	1120	0.0669	0.7771	0.8618	0.9829	0.9909	0.7327	0.9825	0.5717
0.0191	67.06	1140	0.0662	0.7779	0.8696	0.9826	0.9901	0.7490	0.9822	0.5737
0.0123	68.24	1160	0.0668	0.7775	0.8591	0.9830	0.9913	0.7270	0.9826	0.5723
0.0127	69.41	1180	0.0676	0.7772	0.8637	0.9828	0.9907	0.7366	0.9824	0.5720
0.0092	70.59	1200	0.0673	0.7778	0.8699	0.9826	0.9901	0.7496	0.9822	0.5735
0.0101	71.76	1220	0.0680	0.7761	0.8694	0.9824	0.9899	0.7489	0.9820	0.5703
0.0204	72.94	1240	0.0676	0.7772	0.8640	0.9828	0.9907	0.7373	0.9824	0.5721
0.008	74.12	1260	0.0685	0.7768	0.8661	0.9826	0.9904	0.7417	0.9822	0.5714
0.0124	75.29	1280	0.0676	0.7776	0.8648	0.9828	0.9907	0.7390	0.9824	0.5729
0.0134	76.47	1300	0.0689	0.7770	0.8672	0.9826	0.9903	0.7441	0.9822	0.5718
0.0082	77.65	1320	0.0688	0.7755	0.8621	0.9826	0.9907	0.7336	0.9822	0.5687
0.0125	78.82	1340	0.0698	0.7761	0.8655	0.9826	0.9904	0.7407	0.9822

📄 License

The model is released under the other license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご