SegFormer B0 Open-source Tooth Segmentation Model - Free Deployment for Accurate Segmentation of Tooth Regions in Dental X-ray Images

Segformer B0 Finetuned Teeth Segmentation

Developed by vimassaru

A dental X-ray image segmentation model fine-tuned based on the MIT-B0 architecture, specifically designed for precise segmentation of tooth regions in dental imaging

Image Segmentation

Transformers

Open Source License:Other #Dental X-ray Segmentation #High-precision Tooth Identification #Medical Image Analysis

Downloads 55

Release Time : 7/4/2023

Model Overview

This model is a semantic segmentation model optimized for dental X-ray images, capable of identifying and segmenting individual tooth regions in the images, supporting the classification of 32 teeth according to the international standard tooth numbering system

Model Features

High-precision Tooth Segmentation

Achieves an average IoU of 0.73 and an overall accuracy of 0.81 on the test set, with segmentation accuracy exceeding 85% for key teeth (such as incisors and molars)

Complete Tooth Numbering Support

Supports the international standard tooth numbering system (numbers 11-48), capable of identifying 32 individual teeth

Lightweight Architecture

Based on the lightweight design of SegFormer-B0, suitable for real-time processing requirements in medical scenarios

Model Capabilities

X-ray Image Analysis

Tooth Region Segmentation

Multi-class Semantic Segmentation

Medical Image Processing

Use Cases

Dental Diagnosis

Tooth Localization Analysis

Automatically identifies the precise location and boundaries of each tooth in X-ray images

Average localization accuracy of 81.84%

Tooth Health Assessment

Works with other algorithms to detect caries or analyze periodontal disease

Dental Education

Teaching Assistance Annotation

Automatically generates tooth annotations for teaching demonstrations

🚀 segformer-b0-finetuned-segments-sidewalk-oct-22

This model is a fine - tuned version of [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0) on the vimassaru/teethsegmentation dataset, aiming to solve image segmentation tasks in the field of odontology using X - ray images.

🚀 Quick Start

This fine - tuned model segformer - b0 - finetuned - segments - sidewalk - oct - 22 is based on the pre - trained model [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0) and trained on the vimassaru/teethsegmentation dataset. It can be used for image segmentation tasks in the field of odontology, especially for X - ray images.

Evaluation Results

It achieves the following results on the evaluation set:

Loss: 0.1880
Mean Iou: 0.7311
Mean Accuracy: 0.8106
Overall Accuracy: 0.8184
Accuracy Background: nan
Accuracy 11: 0.8456
Accuracy 12: 0.8351
Accuracy 13: 0.8619
Accuracy 14: 0.8112
Accuracy 15: 0.8087
Accuracy 16: 0.9141
Accuracy 17: 0.8742
Accuracy 18: 0.7394
Accuracy 21: 0.8758
Accuracy 22: 0.8579
Accuracy 23: 0.8480
Accuracy 24: 0.7169
Accuracy 25: 0.8273
Accuracy 26: 0.8481
Accuracy 27: 0.8284
Accuracy 28: 0.7298
Accuracy 31: 0.7495
Accuracy 32: 0.7987
Accuracy 33: 0.8661
Accuracy 34: 0.8392
Accuracy 35: 0.7596
Accuracy 36: 0.7482
Accuracy 37: 0.8109
Accuracy 38: 0.7016
Accuracy 41: 0.7217
Accuracy 42: 0.7480
Accuracy 43: 0.8447
Accuracy 44: 0.7868
Accuracy 45: 0.8250
Accuracy 46: 0.8762
Accuracy 47: 0.8519
Accuracy 48: 0.7878
Iou Background: 0.0
Iou 11: 0.8226
Iou 12: 0.8155
Iou 13: 0.8048
Iou 14: 0.7807
Iou 15: 0.7909
Iou 16: 0.8609
Iou 17: 0.8145
Iou 18: 0.6999
Iou 21: 0.8266
Iou 22: 0.8160
Iou 23: 0.8000
Iou 24: 0.6900
Iou 25: 0.7760
Iou 26: 0.8065
Iou 27: 0.7338
Iou 28: 0.6771
Iou 31: 0.6604
Iou 32: 0.7394
Iou 33: 0.7977
Iou 34: 0.7577
Iou 35: 0.6944
Iou 36: 0.6774
Iou 37: 0.7224
Iou 38: 0.6099
Iou 41: 0.6166
Iou 42: 0.6741
Iou 43: 0.7706
Iou 44: 0.7386
Iou 45: 0.7555
Iou 46: 0.8271
Iou 47: 0.8210
Iou 48: 0.7466

🔧 Technical Details

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0006
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 200

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy 11	Accuracy 12	Accuracy 13	Accuracy 14	Accuracy 15	Accuracy 16	Accuracy 17	Accuracy 18	Accuracy 21	Accuracy 22	Accuracy 23	Accuracy 24	Accuracy 25	Accuracy 26	Accuracy 27	Accuracy 28	Accuracy 31	Accuracy 32	Accuracy 33	Accuracy 34	Accuracy 35	Accuracy 36	Accuracy 37	Accuracy 38	Accuracy 41	Accuracy 42	Accuracy 43	Accuracy 44	Accuracy 45	Accuracy 46	Accuracy 47	Accuracy 48	Iou Background	Iou 11	Iou 12	Iou 13	Iou 14	Iou 15	Iou 16	Iou 17	Iou 18	Iou 21	Iou 22	Iou 23	Iou 24	Iou 25	Iou 26	Iou 27	Iou 28	Iou 31	Iou 32	Iou 33	Iou 34	Iou 35	Iou 36	Iou 37	Iou 38	Iou 41	Iou 42	Iou 43	Iou 44	Iou 45	Iou 46	Iou 47	Iou 48
1.5068	2.0	20	1.5257	0.0331	0.0872	0.1082	nan	0.0	0.0	0.0325	0.0	0.0	0.3861	0.7782	0.0	0.5551	0.0	0.0003	0.0	0.0	0.0025	0.5879	0.0	0.0	0.0	0.0	0.0	0.0	0.1233	0.1058	0.0	0.0	0.0	0.0001	0.0	0.0	0.0	0.2200	0.0	0.0	0.0	0.0	0.0298	0.0	0.0	0.2256	0.2666	0.0	0.0733	0.0	0.0003	0.0	0.0	0.0025	0.1063	0.0	0.0	0.0	0.0	0.0	0.0	0.0983	0.0956	0.0	0.0	0.0	0.0001	0.0	0.0	0.0	0.1930	0.0
0.6989	4.0	40	0.6959	0.0319	0.0406	0.0525	nan	0.0032	0.0	0.0	0.0	0.0	0.1235	0.5046	0.0	0.0025	0.0	0.0023	0.0	0.0	0.0821	0.2861	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0411	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2543	0.0	0.0	0.0032	0.0	0.0	0.0	0.0	0.1124	0.3456	0.0	0.0024	0.0	0.0023	0.0	0.0	0.0706	0.2383	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0395	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2387	0.0
0.5093	6.0	60	0.4954	0.1659	0.2494	0.3064	nan	0.4165	0.0	0.5606	0.2388	0.2757	0.7236	0.7757	0.0	0.6343	0.0	0.5877	0.0009	0.0374	0.7495	0.7771	0.0	0.0	0.0	0.0	0.0323	0.0	0.0006	0.8019	0.0	0.0	0.0	0.0	0.0436	0.0	0.0556	0.7505	0.5182	0.0	0.2822	0.0	0.3631	0.2026	0.2067	0.4694	0.5250	0.0	0.3620	0.0	0.4125	0.0009	0.0358	0.4261	0.5167	0.0	0.0	0.0	0.0	0.0299	0.0	0.0006	0.5306	0.0	0.0	0.0	0.0	0.0423	0.0	0.0555	0.5479	0.4643
0.3942	8.0	80	0.3680	0.3501	0.4658	0.5243	nan	0.5738	0.1433	0.7262	0.5454	0.5830	0.7900	0.8112	0.0005	0.7442	0.0533	0.8308	0.3071	0.5064	0.8316	0.7153	0.0	0.0000	0.0743	0.6079	0.4475	0.1634	0.6588	0.7590	0.2711	0.0	0.3319	0.4395	0.6430	0.2279	0.7393	0.6828	0.6962	0.0	0.4491	0.1389	0.5260	0.4596	0.5023	0.6223	0.5467	0.0005	0.5418	0.0531	0.5015	0.2715	0.4175	0.6065	0.4828	0.0	0.0000	0.0713	0.3230	0.3481	0.1586	0.5433	0.5740	0.2533	0.0	0.2423	0.3966	0.4280	0.2200	0.6234	0.6177	0.6324
0.2784	10.0	100	0.2889	0.4636	0.5684	0.6056	nan	0.7515	0.3831	0.6687	0.4978	0.5445	0.7171	0.8213	0.3704	0.8125	0.4238	0.7137	0.3553	0.3746	0.6771	0.7124	0.3304	0.2737	0.3675	0.6784	0.5418	0.5543	0.6343	0.7606	0.5879	0.0628	0.4605	0.6789	0.6545	0.4182	0.8261	0.8022	0.7318	0.0	0.5755	0.3346	0.5240	0.4448	0.5151	0.6337	0.6315	0.3573	0.5967	0.3743	0.5367	0.2860	0.3354	0.6064	0.5735	0.3130	0.2389	0.3153	0.5370	0.4762	0.4740	0.5568	0.6436	0.4934	0.0622	0.3368	0.5307	0.4908	0.4042	0.7224	0.7127	0.6656
0.2464	12.0	120	0.2468	0.5400	0.6505	0.6768	nan	0.7948	0.5458	0.7406	0.5603	0.6560	0.8367	0.8329	0.4258	0.8497	0.5678	0.7320	0.5367	0.6036	0.7622	0.7365	0.5251	0.4648	0.6275	0.7353	0.5512	0.6709	0.6991	0.7896	0.6117	0.2989	0.5592	0.6312	0.5789	0.6009	0.8589	0.7659	0.6656	0.0	0.6477	0.4822	0.5850	0.4886	0.6094	0.7392	0.6728	0.4202	0.6684	0.4945	0.6069	0.4492	0.4973	0.6686	0.6208	0.4915	0.3555	0.4906	0.5972	0.4926	0.5760	0.6233	0.6776	0.5277	0.2650	0.4418	0.5141	0.4982	0.5567	0.7441	0.6964	0.6216
0.1842	14.0	140	0.2142	0.6089	0.7198	0.7422	nan	0.7984	0.6803	0.7803	0.6313	0.7672	0.8662	0.8918	0.6084	0.8619	0.7294	0.8191	0.6076	0.7007	0.8122	0.8322	0.3832	0.5565	0.6651	0.8472	0.7226	0.7371	0.7750	0.7743	0.5362	0.5420	0.6400	0.7419	0.7320	0.6402	0.8401	0.8309	0.6816	0.0	0.7143	0.6368	0.7007	0.5825	0.6365	0.7816	0.7521	0.5826	0.7414	0.6622	0.7151	0.5453	0.5588	0.7077	0.6385	0.3652	0.4556	0.5434	0.7138	0.6374	0.6336	0.6464	0.6580	0.4844	0.4542	0.5349	0.6565	0.6279	0.5893	0.7605	0.7330	0.6423
0.1488	16.0	160	0.1933	0.6322	0.7373	0.7542	nan	0.7908	0.6508	0.7861	0.7352	0.7521	0.8892	0.8725	0.6348	0.8581	0.6927	0.8070	0.6642	0.7642	0.8278	0.8363	0.5789	0.6088	0.6556	0.8139	0.6966	0.6829	0.7853	0.7453	0.6678	0.5757	0.6036	0.7537	0.7381	0.7114	0.8446	0.8103	0.7584	0.0	0.6990	0.6113	0.7069	0.6612	0.7097	0.7968	0.7383	0.6079	0.7618	0.6629	0.7209	0.5896	0.6152	0.7371	0.6847	0.5395	0.5130	0.5606	0.6936	0.6153	0.5981	0.6347	0.6535	0.5868	0.4842	0.5324	0.6701	0.6410	0.6346	0.7595	0.7362	0.7052
0.1416	18.0	180	0.1841	0.6533	0.7530	0.7611	nan	0.8485	0.7657	0.8163	0.6475	0.6706	0.7889	0.8295	0.8478	0.8320	0.7455	0.7939	0.5882	0.6691	0.7761	0.7446	0.7167	0.5879	0.6785	0.7867	0.7367	0.7456	0.7522	0.7657	0.7979	0.6336	0.7404	0.8204	0.7517	0.6922	0.8420	0.8470	0.8352	0.0	0.7802	0.7301	0.7232	0.6005	0.6380	0.7526	0.7347	0.6939	0.7603	0.6879	0.6913	0.5221	0.5773	0.7374	0.6773	0.6223	0.5234	0.5861	0.7062	0.6492	0.6365	0.6644	0.6850	0.6211	0.5522	0.6502	0.7557	0.6860	0.6374	0.7636	0.7728	0.7395
0.1346	20.0	200	0.1742	0.6462	0.7375	0.7528	nan	0.8037	0.7862	0.7855	0.7327	0.7617	0.8752	0.8049	0.5045	0.8113	0.7420	0.8055	0.5818	0.6962	0.7828	0.8136	0.5866	0.5741	0.7100	0.8119	0.7525	0.6603	0.7224	0.8027	0.6645	0.6230	0.7702	0.8144	0.6772	0.7818	0.8084	0.8461	0.7052	0.0	0.7651	0.7384	0.7206	0.6790	0.7135	0.7803	0.6876	0.4954	0.7515	0.6922	0.7354	0.5483	0.6050	0.7345	0.6809	0.5512	0.5261	0.6238	0.7292	0.6495	0.5871	0.6396	0.7011	0.5818	0.5478	0.6583	0.7103	0.6303	0.6758	0.7516	0.7592	0.6734
0.1042	22.0	220	0.1637	0.6

📄 License

This 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