Segformer-b0-finetuned-batch1w5-15Dec open-source image segmentation model accurately accomplishes pixel-level classification tasks

Segformer B0 Finetuned Batch1w5 15Dec

Developed by PushkarA07

A lightweight image segmentation model based on the SegFormer architecture, fine-tuned on specific datasets, excelling in high-precision pixel-level classification tasks

Image Segmentation

Transformers

Open Source License:Other #High-precision image segmentation #Industrial defect detection #Lightweight SegFormer

Downloads 15

Release Time : 12/14/2024

Model Overview

This model is a fine-tuned version of SegFormer-B0, specifically designed for image segmentation tasks. It demonstrates excellent intersection-over-union and accuracy metrics in evaluations, with particularly strong performance in anomaly detection.

Model Features

High-precision segmentation

Achieves a mean intersection-over-union of 0.9143 and mean accuracy of 0.9529 on the evaluation set

Anomaly detection advantage

Specially optimized for anomaly detection with an anomaly intersection-over-union of 0.8302

Lightweight architecture

Based on the lightweight SegFormer-B0 design, suitable for resource-constrained environments

Model Capabilities

Image semantic segmentation

Pixel-level classification

Anomaly region detection

High-resolution image processing

Use Cases

Industrial inspection

Product defect detection

Identifies abnormal regions on manufactured product surfaces

Anomaly detection accuracy of 90.66%

Medical imaging

Lesion region segmentation

Marks abnormal tissue regions in medical images

🚀 segformer-b0-finetuned-batch1w5-15Dec

This model is a fine - tuned version of [PushkarA07/segformer - b0 - finetuned - batch2w5 - 15Dec](https://huggingface.co/PushkarA07/segformer - b0 - finetuned - batch2w5 - 15Dec) on the PushkarA07/batch1 - tiles_W5 dataset. It offers high - performance image segmentation capabilities, achieving excellent results on the evaluation set.

🚀 Quick Start

This model is designed for image segmentation tasks. You can use it directly after fine - tuning on the relevant dataset.

📚 Documentation

Model Information

Property	Details
Model Type	segformer - b0 - finetuned - batch1w5 - 15Dec
Training Data	PushkarA07/batch1 - tiles_W5 dataset

Evaluation Results

This model achieves the following results on the evaluation set:

Loss: 0.0038
Mean Iou: 0.9143
Mean Accuracy: 0.9529
Overall Accuracy: 0.9985
Accuracy Abnormality: 0.9066
Iou Abnormality: 0.8302

Training Procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e - 05
train_batch_size: 4
eval_batch_size: 4
seed: 42
optimizer: Use 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: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Abnormality	Iou Abnormality
0.0085	0.8333	10	0.0101	0.8170	0.8872	0.9963	0.7762	0.6376
0.0097	1.6667	20	0.0078	0.8448	0.8888	0.9971	0.7787	0.6926
0.0102	2.5	30	0.0071	0.8563	0.9028	0.9973	0.8068	0.7153
0.0062	3.3333	40	0.0066	0.8618	0.9005	0.9975	0.8018	0.7262
0.006	4.1667	50	0.0062	0.8693	0.9147	0.9976	0.8304	0.7410
0.0086	5.0	60	0.0060	0.8726	0.9194	0.9976	0.8398	0.7475
0.0056	5.8333	70	0.0056	0.8773	0.9128	0.9978	0.8264	0.7568
0.0044	6.6667	80	0.0056	0.8789	0.9270	0.9978	0.8550	0.7601
0.0045	7.5	90	0.0054	0.8818	0.9239	0.9978	0.8487	0.7658
0.0087	8.3333	100	0.0053	0.8850	0.9347	0.9979	0.8705	0.7721
0.0045	9.1667	110	0.0052	0.8835	0.9150	0.9979	0.8306	0.7692
0.0051	10.0	120	0.0051	0.8888	0.9360	0.9980	0.8730	0.7798
0.0045	10.8333	130	0.0049	0.8904	0.9270	0.9980	0.8547	0.7827
0.0068	11.6667	140	0.0048	0.8904	0.9290	0.9980	0.8589	0.7828
0.0029	12.5	150	0.0048	0.8924	0.9394	0.9980	0.8799	0.7867
0.0051	13.3333	160	0.0048	0.8943	0.9361	0.9981	0.8731	0.7906
0.0038	14.1667	170	0.0047	0.8953	0.9394	0.9981	0.8796	0.7926
0.0075	15.0	180	0.0047	0.8967	0.9416	0.9981	0.8841	0.7952
0.0054	15.8333	190	0.0047	0.8954	0.9315	0.9981	0.8637	0.7928
0.0031	16.6667	200	0.0046	0.8973	0.9373	0.9981	0.8755	0.7965
0.0049	17.5	210	0.0046	0.8970	0.9300	0.9982	0.8606	0.7958
0.0049	18.3333	220	0.0045	0.9001	0.9430	0.9982	0.8870	0.8019
0.0038	19.1667	230	0.0045	0.9002	0.9485	0.9982	0.8979	0.8022
0.0074	20.0	240	0.0045	0.9009	0.9424	0.9982	0.8856	0.8036
0.0048	20.8333	250	0.0045	0.9008	0.9473	0.9982	0.8955	0.8034
0.0058	21.6667	260	0.0045	0.9011	0.9464	0.9982	0.8938	0.8039
0.0051	22.5	270	0.0044	0.9029	0.9421	0.9983	0.8850	0.8075
0.0062	23.3333	280	0.0043	0.9026	0.9379	0.9983	0.8766	0.8070
0.0051	24.1667	290	0.0044	0.9027	0.9440	0.9982	0.8888	0.8071
0.0026	25.0	300	0.0043	0.9043	0.9443	0.9983	0.8894	0.8103
0.007	25.8333	310	0.0043	0.9042	0.9498	0.9983	0.9004	0.8102
0.0041	26.6667	320	0.0043	0.9046	0.9454	0.9983	0.8916	0.8110
0.0045	27.5	330	0.0043	0.9048	0.9427	0.9983	0.8862	0.8114
0.0041	28.3333	340	0.0043	0.9055	0.9490	0.9983	0.8988	0.8128
0.0024	29.1667	350	0.0042	0.9064	0.9485	0.9983	0.8979	0.8145
0.0035	30.0	360	0.0042	0.9061	0.9424	0.9983	0.8856	0.8139
0.003	30.8333	370	0.0042	0.9063	0.9523	0.9983	0.9056	0.8142
0.0054	31.6667	380	0.0042	0.9074	0.9447	0.9983	0.8902	0.8165
0.0054	32.5	390	0.0042	0.9064	0.9480	0.9983	0.8969	0.8144
0.0041	33.3333	400	0.0042	0.9053	0.9471	0.9983	0.8951	0.8123
0.0059	34.1667	410	0.0041	0.9075	0.9439	0.9983	0.8886	0.8166
0.0027	35.0	420	0.0042	0.9066	0.9452	0.9983	0.8912	0.8149
0.0052	35.8333	430	0.0042	0.9074	0.9474	0.9983	0.8956	0.8165
0.0042	36.6667	440	0.0041	0.9070	0.9457	0.9983	0.8922	0.8156
0.0037	37.5	450	0.0041	0.9076	0.9457	0.9983	0.8922	0.8170
0.0033	38.3333	460	0.0041	0.9084	0.9481	0.9984	0.8970	0.8185
0.0031	39.1667	470	0.0041	0.9085	0.9471	0.9984	0.8949	0.8187
0.0037	40.0	480	0.0042	0.9071	0.9543	0.9983	0.9096	0.8159
0.0048	40.8333	490	0.0041	0.9088	0.9500	0.9984	0.9008	0.8192
0.0042	41.6667	500	0.0041	0.9086	0.9474	0.9984	0.8957	0.8188
0.0024	42.5	510	0.0040	0.9095	0.9470	0.9984	0.8948	0.8206
0.0047	43.3333	520	0.0040	0.9091	0.9511	0.9984	0.9031	0.8198
0.0054	44.1667	530	0.0041	0.9080	0.9438	0.9984	0.8884	0.8176
0.0053	45.0	540	0.0041	0.9084	0.9460	0.9984	0.8928	0.8185
0.0033	45.8333	550	0.0041	0.9094	0.9515	0.9984	0.9038	0.8205
0.0044	46.6667	560	0.0042	0.9076	0.9580	0.9983	0.9171	0.8169
0.0021	47.5	570	0.0040	0.9095	0.9501	0.9984	0.9010	0.8206
0.0035	48.3333	580	0.0040	0.9092	0.9529	0.9983	0.9067	0.8200
0.0038	49.1667	590	0.0040	0.9109	0.9505	0.9984	0.9019	0.8234
0.004	50.0	600	0.0041	0.9103	0.9563	0.9984	0.9134	0.8223
0.0044	50.8333	610	0.0040	0.9106	0.9464	0.9984	0.8936	0.8229
0.0026	51.6667	620	0.0040	0.9104	0.9554	0.9984	0.9116	0.8225
0.0062	52.5	630	0.0040	0.9114	0.9510	0.9984	0.9027	0.8244
0.0023	53.3333	640	0.0040	0.9114	0.9470	0.9984	0.8948	0.8244
0.0029	54.1667	650	0.0040	0.9113	0.9508	0.9984	0.9024	0.8242
0.0042	55.0	660	0.0040	0.9116	0.9528	0.9984	0.9064	0.8248
0.0044	55.8333	670	0.0039	0.9121	0.9519	0.9984	0.9045	0.8258
0.0016	56.6667	680	0.0040	0.9116	0.9514	0.9984	0.9035	0.8248
0.0044	57.5	690	0.0039	0.9116	0.9533	0.9984	0.9075	0.8248
0.0031	58.3333	700	0.0039	0.9118	0.9497	0.9984	0.9002	0.8253
0.0038	59.1667	710	0.0039	0.9119	0.9509	0.9984	0.9025	0.8254
0.0042	60.0	720	0.0040	0.9117	0.9535	0.9984	0.9078	0.8250
0.0045	60.8333	730	0.0039	0.9119	0.9512	0.9984	0.9032	0.8254
0.0039	61.6667	740	0.0039	0.9122	0.9507	0.9984	0.9022	0.8260
0.0022	62.5	750	0.0040	0.9117	0.9562	0.9984	0.9134	0.8250
0.0039	63.3333	760	0.0039	0.9126	0.9502	0.9984	0.9012	0.8268
0.0031	64.1667	770	0.0039	0.9115	0.9507	0.9984	0.9021	0.8245
0.0037	65.0	780	0.0040	0.9118	0.9533	0.9984	0.9074	0.8252
0.0046	65.8333	790	0.0039	0.9123	0.9489	0.9984	0.8986	0.8261
0.0026	66.6667	800	0.0039	0.9127	0.9532	0.9984	0.9073	0.8269
0.0039	67.5	810	0.0039	0.9121	0.9469	0.9984	0.8928	0.8263
0.0034	68.3333	820	0.0039	0.9126	0.9523	0.9984	0.9056	0.8270
0.0032	69.1667	830	0.0039	0.9127	0.9508	0.9984	0.9024	0.8271
0.0038	70.0	840	0.0039	0.9128	0.9529	0.9984	0.9067	0.8272
0.0034	70.8333	850	0.0039	0.9129	0.9535	0.9984	0.9078	0.8273
0.0038	71.6667	860	0.0039	0.9130	0.9512	0.9984	0.9032	0.8274
0.0039	72.5	870	0.0039	0.9131	0.9507	0.9984	0.9022	0.8275
0.0022	73.3333	880	0.0040	0.9127	0.9562	0.9984	0.9134	0.8270
0.0039	74.1667	890	0.0039	0.9132	0.9502	0.9984	0.9012	0.8276
0.0031	75.0	900	0.0039	0.9125	0.9507	0.9984	0.9021	0.8271
0.0037	75.8333	910	0.0040	0.9128	0.9533	0.9984	0.9074	0.8272
0.0046	76.6667	920	0.0039	0.9133	0.9489	0.9984	0.8986	0.8277
0.0026	77.5	930	0.0039	0.9134	0.9532	0.9984	0.9073	0.8278
0.0039	78.3333	940	0.0039	0.9131	0.9469	0.9984	0.8928	0.8273
0.0034	79.1667	950	0.0039	0.9132	0.9523	0.9984	0.9056	0.8279
0.0032	80.0	960	0.0039	0.9133	0.9508	0.9984	0.9024	0.8280
0.0038	80.8333	970	0.0039	0.9134	0.9529	0.9984	0.9067	0.8281
0.0034	81.6667	980	0.0039	0.9135	0.9535	0.9984	0.9078	0.8282
0.0038	82.5	990	0.0039	0.9136	0.9512	0.9984	0.9032	0.8283
0.0039	83.3333	1000	0.0039	0.9137	0.9507	0.9984	0.9022	0.8284

📄 License

This model is 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