SegFormer-B0 Open-source Image Segmentation Model - Achieve Excellent Image Segmentation on Specific Datasets after Fine-tuning

Segformer B0 Finetuned Morphpadver1 Hgo Coord

Developed by NICOPOI-9

An image segmentation model fine-tuned based on nvidia/mit-b0, performing excellently on the NICOPOI-9/morphpad_coord_hgo_512_4class dataset

Image Segmentation

Transformers

Open Source License:Other #High-precision image segmentation #Multi-directional morphological analysis #Medical image processing

Downloads 98

Release Time : 4/21/2025

Model Overview

This model is a SegFormer architecture optimized for image segmentation tasks, particularly suitable for direction-sensitive image segmentation scenarios

Model Features

High-precision directional segmentation

Performs excellently in four directions: 0-0, 0-90, 90-0, and 90-90, with an average IoU of 0.9858

Small model optimization

Based on the lightweight SegFormer-B0 architecture, suitable for deployment in resource-constrained environments

Domain-specific optimization

Specially optimized for the NICOPOI-9/morphpad_coord_hgo_512_4class dataset

Model Capabilities

Image segmentation

Direction-sensitive analysis

High-precision pixel-level classification

Use Cases

Medical image analysis

Tissue segmentation

Used for precise segmentation of specific tissues in medical images

Industrial inspection

Defect detection

Detects and precisely locates surface defects on products

🚀 segformer-b0-finetuned-morphpadver1-hgo-coord

This model is a fine - tuned version of [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0) on the NICOPOI - 9/morphpad_coord_hgo_512_4class dataset. It offers high - performance results in image segmentation tasks, achieving excellent scores on multiple evaluation metrics.

🚀 Quick Start

Evaluation Results

This model achieves the following results on the evaluation set:

Loss: 0.0306
Mean Iou: 0.9858
Mean Accuracy: 0.9928
Overall Accuracy: 0.9928
Accuracy 0 - 0: 0.9933
Accuracy 0 - 90: 0.9937
Accuracy 90 - 0: 0.9943
Accuracy 90 - 90: 0.9898
Iou 0 - 0: 0.9885
Iou 0 - 90: 0.9850
Iou 90 - 0: 0.9826
Iou 90 - 90: 0.9872

📚 Documentation

Training and Evaluation Data

More information needed

Training Procedure

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
1.2185	2.5445	4000	1.2349	0.2290	0.3745	0.3762	0.2785	0.4062	0.4936	0.3198	0.2085	0.2334	0.2525	0.2216
1.0978	5.0891	8000	1.1020	0.2905	0.4487	0.4508	0.3780	0.5302	0.5341	0.3524	0.2937	0.2870	0.2991	0.2822
0.9886	7.6336	12000	1.0139	0.3231	0.4871	0.4896	0.4154	0.4500	0.7245	0.3585	0.3291	0.3272	0.3266	0.3096
0.9358	10.1781	16000	0.9575	0.3517	0.5195	0.5215	0.3765	0.6411	0.5865	0.4740	0.3438	0.3539	0.3617	0.3473
0.8735	12.7226	20000	0.8853	0.4007	0.5704	0.5726	0.4998	0.5637	0.7536	0.4647	0.4109	0.3953	0.4055	0.3913
0.7186	15.2672	24000	0.6833	0.5558	0.7151	0.7141	0.7389	0.6650	0.6919	0.7647	0.5919	0.5261	0.5453	0.5598
0.6514	17.8117	28000	0.4379	0.7017	0.8243	0.8243	0.8344	0.8161	0.8279	0.8187	0.7198	0.6807	0.6933	0.7130
0.603	20.3562	32000	0.2900	0.7980	0.8879	0.8874	0.9117	0.8490	0.8888	0.9020	0.8160	0.7726	0.7893	0.8142
0.2448	22.9008	36000	0.2154	0.8496	0.9184	0.9185	0.9330	0.9179	0.9170	0.9058	0.8683	0.8329	0.8445	0.8527
0.2766	25.4453	40000	0.2004	0.8612	0.9254	0.9254	0.9487	0.9059	0.9381	0.9088	0.8717	0.8469	0.8635	0.8628
0.6278	27.9898	44000	0.1410	0.8976	0.9459	0.9459	0.9426	0.9377	0.9559	0.9474	0.9075	0.8863	0.8932	0.9034
0.1684	30.5344	48000	0.1163	0.9137	0.9549	0.9548	0.9595	0.9417	0.9579	0.9605	0.9245	0.9046	0.9069	0.9187
0.0638	33.0789	52000	0.0927	0.9338	0.9657	0.9657	0.9697	0.9589	0.9715	0.9627	0.9406	0.9291	0.9291	0.9363
0.0749	35.6234	56000	0.0836	0.9382	0.9680	0.9680	0.9714	0.9663	0.9680	0.9664	0.9449	0.9325	0.9339	0.9414
0.045	38.1679	60000	0.0624	0.9545	0.9767	0.9767	0.9787	0.9751	0.9763	0.9766	0.9587	0.9521	0.9499	0.9573
0.1278	40.7125	64000	0.0635	0.9546	0.9767	0.9767	0.9773	0.9743	0.9813	0.9737	0.9598	0.9521	0.9492	0.9572
0.0443	43.2570	68000	0.0598	0.9584	0.9787	0.9787	0.9815	0.9723	0.9858	0.9752	0.9624	0.9548	0.9548	0.9617
0.0337	45.8015	72000	0.0549	0.9622	0.9807	0.9807	0.9877	0.9804	0.9820	0.9726	0.9648	0.9587	0.9622	0.9632
0.0434	48.3461	76000	0.0539	0.9643	0.9816	0.9817	0.9793	0.9779	0.9913	0.9781	0.9691	0.9611	0.9565	0.9703
0.1576	50.8906	80000	0.0577	0.9656	0.9825	0.9825	0.9799	0.9822	0.9825	0.9856	0.9694	0.9634	0.9653	0.9645
0.025	53.4351	84000	0.0453	0.9724	0.9860	0.9860	0.9856	0.9884	0.9840	0.9858	0.9762	0.9698	0.9697	0.9739
0.0318	55.9796	88000	0.0401	0.9733	0.9865	0.9865	0.9884	0.9845	0.9865	0.9865	0.9766	0.9700	0.9714	0.9753
0.1355	58.5242	92000	0.0453	0.9764	0.9880	0.9880	0.9896	0.9874	0.9889	0.9861	0.9796	0.9742	0.9731	0.9786
0.0256	61.0687	96000	0.0359	0.9817	0.9907	0.9908	0.9902	0.9925	0.9902	0.9901	0.9846	0.9808	0.9783	0.9833
0.019	63.6132	100000	0.0320	0.9819	0.9908	0.9909	0.9914	0.9908	0.9936	0.9875	0.9838	0.9812	0.9787	0.9841
0.0713	66.1578	104000	0.0319	0.9827	0.9912	0.9912	0.9940	0.9922	0.9937	0.9847	0.9859	0.9812	0.9807	0.9828
0.1036	68.7023	108000	0.0369	0.9807	0.9902	0.9903	0.9932	0.9916	0.9946	0.9813	0.9844	0.9807	0.9790	0.9788
0.0575	71.2468	112000	0.0338	0.9843	0.9921	0.9921	0.9939	0.9913	0.9929	0.9901	0.9870	0.9822	0.9814	0.9867
0.0136	73.7913	116000	0.0259	0.9870	0.9934	0.9934	0.9926	0.9936	0.9946	0.9930	0.9889	0.9852	0.9850	0.9891
0.045	76.3359	120000	0.0310	0.9844	0.9921	0.9921	0.9913	0.9926	0.9941	0.9902	0.9866	0.9834	0.9805	0.9871
0.6665	78.8804	124000	0.0306	0.9858	0.9928	0.9928	0.9933	0.9937	0.9943	0.9898	0.9885	0.9850	0.9826	0.9872

Framework Versions

Transformers 4.48.3
Pytorch 2.1.0
Datasets 3.2.0
Tokenizers 0.21.0

📄 License

The license for 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