SegFormer-B0 Finetuned Morphpadver1-HGO-3 Open-source Image Segmentation Model - Achieve High-precision Image Segmentation for Free

Segformer B0 Finetuned Morphpadver1 Hgo 3

Developed by NICOPOI-9

An image segmentation model fine-tuned based on nvidia/mit-b0, trained on the NICOPOI-9/morphpad_hgo_512_4class dataset, excelling in high-precision image segmentation tasks.

Image Segmentation

Transformers

Open Source License:Other #High-precision image segmentation #Medical image analysis #MIT-B0 fine-tuning

Downloads 30

Release Time : 2/21/2025

Model Overview

This model is a lightweight implementation of the SegFormer architecture, specifically designed for image segmentation tasks. Fine-tuned on a specific dataset, it can accurately identify and segment different regions in images.

Model Features

High-precision segmentation

Achieves an average IoU of 0.978 and an average accuracy of 0.9889 on the evaluation set, demonstrating excellent performance.

Lightweight architecture

Based on the SegFormer-B0 architecture, it reduces computational resource requirements while maintaining performance.

Multi-angle adaptation

The model performs stably on images from different angles (0-0, 0-90, 90-0, 90-90), with minimal accuracy variations.

Model Capabilities

Image segmentation

Pixel-level classification

Multi-category recognition

Use Cases

Medical image analysis

Tissue segmentation

Used for identifying and segmenting specific tissues in medical images

Industrial inspection

Defect detection

Identifies defective areas on product surfaces

🚀 segformer-b0-finetuned-morphpadver1-hgo-3

This is a fine - tuned model based on nvidia/mit-b0, designed for image segmentation tasks on the NICOPOI - 9/morphpad_hgo_512_4class dataset.

🚀 Quick Start

This model is a fine - tuned version of nvidia/mit-b0 on the NICOPOI-9/morphpad_hgo_512_4class dataset. It achieves the following results on the evaluation set:

Loss: 0.0870
Mean Iou: 0.9780
Mean Accuracy: 0.9889
Overall Accuracy: 0.9888
Accuracy 0 - 0: 0.9913
Accuracy 0 - 90: 0.9886
Accuracy 90 - 0: 0.9892
Accuracy 90 - 90: 0.9863
Iou 0 - 0: 0.9811
Iou 0 - 90: 0.9785
Iou 90 - 0: 0.9754
Iou 90 - 90: 0.9771

📚 Documentation

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e - 05
train_batch_size: 1
eval_batch_size: 1
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon = 1e - 08 and optimizer_args = No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 80

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy 0 - 0	Accuracy 0 - 90	Accuracy 90 - 0	Accuracy 90 - 90	Iou 0 - 0	Iou 0 - 90	Iou 90 - 0	Iou 90 - 90
1.2394	2.5445	4000	1.2392	0.2273	0.3754	0.3738	0.4516	0.3985	0.2299	0.4215	0.2520	0.2205	0.1802	0.2565
1.0858	5.0891	8000	1.0986	0.2889	0.4462	0.4463	0.4510	0.3646	0.5196	0.4498	0.3208	0.2453	0.2792	0.3104
1.0059	7.6336	12000	1.0060	0.3295	0.4934	0.4935	0.4793	0.6104	0.4583	0.4256	0.3472	0.3209	0.3057	0.3442
0.9457	10.1781	16000	0.9497	0.3565	0.5236	0.5240	0.4976	0.6276	0.5356	0.4335	0.3777	0.3459	0.3490	0.3532
0.93	12.7226	20000	0.9072	0.3759	0.5397	0.5412	0.4649	0.6012	0.6437	0.4491	0.4073	0.3550	0.3549	0.3863
0.8403	15.2672	24000	0.8349	0.4263	0.5967	0.5959	0.6497	0.5299	0.6555	0.5517	0.4548	0.4083	0.3976	0.4443
0.8826	17.8117	28000	0.6734	0.5172	0.6784	0.6787	0.6590	0.7037	0.6902	0.6606	0.5517	0.4864	0.4837	0.5469
0.8028	20.3562	32000	0.4582	0.6843	0.8116	0.8120	0.7923	0.7953	0.8322	0.8268	0.6986	0.6686	0.6646	0.7056
0.8461	22.9008	36000	0.2845	0.8071	0.8931	0.8928	0.9089	0.8839	0.8825	0.8971	0.8322	0.7928	0.7850	0.8182
0.7034	25.4453	40000	0.3351	0.7724	0.8716	0.8713	0.8859	0.8680	0.8485	0.8839	0.7928	0.7649	0.7472	0.7849
1.0428	27.9898	44000	0.1882	0.8750	0.9334	0.9331	0.9467	0.9277	0.9323	0.9267	0.8912	0.8673	0.8647	0.8767
0.3497	30.5344	48000	0.1620	0.8982	0.9464	0.9461	0.9582	0.9539	0.9279	0.9455	0.9163	0.8878	0.8840	0.9046
0.0803	33.0789	52000	0.1314	0.9187	0.9577	0.9575	0.9711	0.9517	0.9508	0.9574	0.9300	0.9101	0.9121	0.9224
0.1394	35.6234	56000	0.1271	0.9228	0.9598	0.9597	0.9648	0.9612	0.9530	0.9603	0.9332	0.9163	0.9167	0.9250
0.0579	38.1679	60000	0.1170	0.9351	0.9665	0.9664	0.9723	0.9652	0.9619	0.9665	0.9446	0.9300	0.9282	0.9377
0.1097	40.7125	64000	0.1121	0.9402	0.9690	0.9691	0.9634	0.9748	0.9732	0.9644	0.9477	0.9380	0.9334	0.9416
0.0615	43.2570	68000	0.1069	0.9458	0.9720	0.9721	0.9707	0.9713	0.9777	0.9685	0.9523	0.9424	0.9426	0.9459
0.0425	45.8015	72000	0.0967	0.9540	0.9764	0.9764	0.9789	0.9740	0.9798	0.9729	0.9609	0.9506	0.9514	0.9529
0.0396	48.3461	76000	0.0991	0.9599	0.9795	0.9795	0.9827	0.9766	0.9783	0.9805	0.9656	0.9562	0.9560	0.9617
0.4319	50.8906	80000	0.0975	0.9583	0.9786	0.9787	0.9759	0.9826	0.9769	0.9792	0.9607	0.9566	0.9551	0.9608
0.0299	53.4351	84000	0.0959	0.9662	0.9828	0.9827	0.9858	0.9837	0.9807	0.9811	0.9712	0.9655	0.9637	0.9644
0.0282	55.9796	88000	0.0933	0.9687	0.9841	0.9841	0.9866	0.9830	0.9823	0.9846	0.9729	0.9665	0.9655	0.9698
0.1038	58.5242	92000	0.0864	0.9707	0.9852	0.9851	0.9882	0.9874	0.9842	0.9809	0.9727	0.9709	0.9698	0.9695
0.0246	61.0687	96000	0.0990	0.9722	0.9859	0.9859	0.9885	0.9873	0.9847	0.9832	0.9764	0.9716	0.9706	0.9703
0.0208	63.6132	100000	0.0839	0.9749	0.9873	0.9873	0.9901	0.9897	0.9850	0.9845	0.9781	0.9748	0.9726	0.9743
0.0749	66.1578	104000	0.0874	0.9761	0.9879	0.9879	0.9901	0.9888	0.9875	0.9851	0.9798	0.9757	0.9734	0.9753
0.0798	68.7023	108000	0.0866	0.9753	0.9875	0.9875	0.9908	0.9881	0.9883	0.9830	0.9792	0.9759	0.9722	0.9740
0.1236	71.2468	112000	0.0933	0.9775	0.9886	0.9886	0.9911	0.9895	0.9870	0.9868	0.9813	0.9773	0.9748	0.9765
0.0165	73.7913	116000	0.0894	0.9785	0.9891	0.9891	0.9902	0.9912	0.9867	0.9886	0.9818	0.9782	0.9759	0.9783
0.0204	76.3359	120000	0.0885	0.9792	0.9895	0.9895	0.9909	0.9913	0.9874	0.9884	0.9823	0.9791	0.9766	0.9788
0.7823	78.8804	124000	0.0870	0.9780	0.9889	0.9888	0.9913	0.9886	0.9892	0.9863	0.9811	0.9785	0.9754	0.9771

Framework versions

Transformers 4.48.3
Pytorch 2.1.0
Datasets 3.2.0
Tokenizers 0.21.0

📄 License

The license of this model is other.

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