Segformer-b1-solarModuleAnomaly-v0.1 Open-source Image Segmentation Model - Precise Detection of Solar Module Anomalies

Home

Segformer B1 Solarmoduleanomaly V0.1

Developed by zklee98

An image segmentation model fine-tuned on solar module anomaly detection dataset based on nvidia/mit-b1

Image Segmentation

Transformers

Open Source License:Other #Photovoltaic Panel Anomaly Detection #Image Segmentation Model #Industrial Visual Inspection

Downloads 18

Release Time : 4/14/2023

Model Overview

This model is an optimized SegFormer architecture image segmentation model for solar module anomaly detection tasks, capable of identifying abnormal areas in solar modules.

Model Features

Solar Module Anomaly Detection

Specifically optimized for solar module anomaly detection tasks

Efficient Segmentation Architecture

Based on SegFormer-b1 architecture, balancing performance and efficiency

Model Capabilities

Image Segmentation

Anomaly Area Detection

Solar Module Analysis

Use Cases

Energy Industry

Solar Power Plant Inspection

Automatically detect abnormal areas on solar panels

Achieved 0.7643 anomaly detection accuracy on the evaluation set

🚀 segformer-b1-solarModuleAnomaly-v0.1

This model is a fine - tuned version of nvidia/mit-b1 on the zklee98/solarModuleAnomaly dataset. It is designed for image segmentation tasks in the field of vision, offering valuable insights into solar module anomaly detection.

🚀 Quick Start

This model is a fine - tuned version of nvidia/mit-b1 on the zklee98/solarModuleAnomaly dataset. It achieves the following results on the evaluation set:

Loss: 0.1547
Mean Iou: 0.3822
Mean Accuracy: 0.7643
Overall Accuracy: 0.7643
Accuracy Unlabelled: nan
Accuracy Anomaly: 0.7643
Iou Unlabelled: 0.0
Iou Anomaly: 0.7643

🔧 Technical Details

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: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabelled	Accuracy Anomaly	Iou Anomaly
0.4699	0.4	20	0.6337	0.4581	0.9162	0.9162	nan	0.9162	0.9162
0.3129	0.8	40	0.4636	0.3704	0.7407	0.7407	nan	0.7407	0.7407
0.2732	1.2	60	0.3164	0.3867	0.7734	0.7734	nan	0.7734	0.7734
0.2653	1.6	80	0.3769	0.4090	0.8180	0.8180	nan	0.8180	0.8180
0.2232	2.0	100	0.2976	0.2479	0.4958	0.4958	nan	0.4958	0.4958
0.5305	2.4	120	0.3151	0.3807	0.7613	0.7613	nan	0.7613	0.7613
0.2423	2.8	140	0.3189	0.4152	0.8305	0.8305	nan	0.8305	0.8305
0.3341	3.2	160	0.2384	0.3861	0.7723	0.7723	nan	0.7723	0.7723
0.2146	3.6	180	0.3200	0.4621	0.9243	0.9243	nan	0.9243	0.9243
0.1866	4.0	200	0.2510	0.3646	0.7291	0.7291	nan	0.7291	0.7291
0.2861	4.4	220	0.2736	0.4202	0.8404	0.8404	nan	0.8404	0.8404
0.2048	4.8	240	0.2410	0.3912	0.7823	0.7823	nan	0.7823	0.7823
0.1604	5.2	260	0.2233	0.3672	0.7344	0.7344	nan	0.7344	0.7344
0.2756	5.6	280	0.2705	0.4494	0.8987	0.8987	nan	0.8987	0.8987
0.1859	6.0	300	0.2211	0.4045	0.8089	0.8089	nan	0.8089	0.8089
0.1306	6.4	320	0.2140	0.3763	0.7525	0.7525	nan	0.7525	0.7525
0.5508	6.8	340	0.2231	0.4185	0.8371	0.8371	nan	0.8371	0.8371
0.1446	7.2	360	0.2139	0.3666	0.7332	0.7332	nan	0.7332	0.7332
0.3275	7.6	380	0.2470	0.3964	0.7928	0.7928	nan	0.7928	0.7928
0.164	8.0	400	0.2017	0.3910	0.7819	0.7819	nan	0.7819	0.7819
0.1864	8.4	420	0.2307	0.4408	0.8816	0.8816	nan	0.8816	0.8816
0.1578	8.8	440	0.1869	0.3707	0.7414	0.7414	nan	0.7414	0.7414
0.1201	9.2	460	0.2115	0.3834	0.7667	0.7667	nan	0.7667	0.7667
0.1783	9.6	480	0.2009	0.3747	0.7495	0.7495	nan	0.7495	0.7495
0.1232	10.0	500	0.1797	0.3865	0.7729	0.7729	nan	0.7729	0.7729
0.2572	10.4	520	0.1983	0.4057	0.8115	0.8115	nan	0.8115	0.8115
0.1209	10.8	540	0.1607	0.4274	0.8547	0.8547	nan	0.8547	0.8547
0.1234	11.2	560	0.2260	0.4066	0.8133	0.8133	nan	0.8133	0.8133
0.145	11.6	580	0.1963	0.3939	0.7878	0.7878	nan	0.7878	0.7878
0.0665	12.0	600	0.1912	0.3873	0.7747	0.7747	nan	0.7747	0.7747
0.0826	12.4	620	0.2095	0.4186	0.8373	0.8373	nan	0.8373	0.8373
0.1212	12.8	640	0.1732	0.4059	0.8118	0.8118	nan	0.8118	0.8118
0.142	13.2	660	0.2086	0.4007	0.8013	0.8013	nan	0.8013	0.8013
0.0899	13.6	680	0.1838	0.3928	0.7856	0.7856	nan	0.7856	0.7856
0.1857	14.0	700	0.1638	0.4157	0.8315	0.8315	nan	0.8315	0.8315
0.0788	14.4	720	0.1736	0.4112	0.8223	0.8223	nan	0.8223	0.8223
0.2543	14.8	740	0.1547	0.3822	0.7643	0.7643	nan	0.7643	0.7643

Framework versions

Transformers 4.28.0
Pytorch 2.0.0+cu118
Datasets 2.11.0
Tokenizers 0.13.3

📄 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