segformer-b5開源圖像分割模型 - 免費部署助力腳部圖像精準分割處理

首頁

Segformer B5 Finetuned Segments Instryde Foot Test

由plant開發

基於nvidia/mit-b5在inStryde/inStrydeSegmentationFoot數據集上微調的圖像分割模型

圖像分割

Transformers

開源協議:其他 #足部圖像分割 #高精度分割 #醫療影像分析

下載量 19

發布時間 : 8/4/2022

模型概述

該模型是SegFormer架構的B5變體，專門針對足部圖像分割任務進行了微調。適用於醫學影像或足部相關視覺分析場景。

模型特點

高精度分割

在足部圖像分割任務上達到0.9344的總體準確率

專業領域優化

針對醫學/足部影像數據進行了專門微調

高效架構

基於SegFormer的混合Transformer架構，平衡性能與效率

模型能力

足部圖像分割

醫學影像分析

像素級分類

使用案例

醫療影像

足部病變區域分割

識別和分割足部醫學影像中的特定區域

平均IoU達到0.4672

生物識別

足部特徵提取

用於生物特徵識別系統中的足部輪廓分析

🚀 segformer-b5-finetuned-segments-instryde-foot-test

該模型是基於 nvidia/mit-b5 在 inStryde/inStrydeSegmentationFoot 數據集上進行微調的版本。它在評估集上取得了以下效果：

損失：0.0496
平均交併比（Mean Iou）：0.4672
平均準確率（Mean Accuracy）：0.9344
總體準確率（Overall Accuracy）：0.9344
每類交併比（Per Category Iou）：[0.0, 0.9343870058298716]
每類準確率（Per Category Accuracy）：[nan, 0.9343870058298716]

🚀 快速開始

此模型是圖像分割領域的有力工具，基於 nvidia/mit-b5 微調而來，在 inStryde/inStrydeSegmentationFoot 數據集上進行了優化。

📚 詳細文檔

訓練超參數

訓練過程中使用了以下超參數：

學習率（learning_rate）：6e-05
訓練批次大小（train_batch_size）：2
評估批次大小（eval_batch_size）：2
隨機種子（seed）：42
優化器（optimizer）：Adam（β1=0.9，β2=0.999，ε=1e-08）
學習率調度器類型（lr_scheduler_type）：線性
訓練輪數（num_epochs）：50

訓練結果

訓練損失	輪數	步數	驗證損失	平均交併比	平均準確率	總體準確率	每類交併比	每類準確率
0.1392	0.23	20	0.2371	0.4064	0.8128	0.8128	[0.0, 0.8127920708469037]	[nan, 0.8127920708469037]
0.2273	0.45	40	0.0993	0.4449	0.8898	0.8898	[0.0, 0.889800913515142]	[nan, 0.889800913515142]
0.0287	0.68	60	0.0607	0.4190	0.8379	0.8379	[0.0, 0.8379005425233161]	[nan, 0.8379005425233161]
0.03	0.91	80	0.0572	0.4072	0.8144	0.8144	[0.0, 0.8144304164916533]	[nan, 0.8144304164916533]
0.0239	1.14	100	0.0577	0.3973	0.7946	0.7946	[0.0, 0.7946284254068925]	[nan, 0.7946284254068925]
0.0196	1.36	120	0.0425	0.4227	0.8455	0.8455	[0.0, 0.8454754171184029]	[nan, 0.8454754171184029]
0.0295	1.59	140	0.0368	0.4479	0.8958	0.8958	[0.0, 0.895802316554768]	[nan, 0.895802316554768]
0.0297	1.82	160	0.0441	0.4561	0.9121	0.9121	[0.0, 0.9121241975954804]	[nan, 0.9121241975954804]
0.0276	2.05	180	0.0332	0.4629	0.9258	0.9258	[0.0, 0.925774145806165]	[nan, 0.925774145806165]
0.0148	2.27	200	0.0395	0.4310	0.8621	0.8621	[0.0, 0.8620666905637888]	[nan, 0.8620666905637888]
0.012	2.5	220	0.0372	0.4381	0.8761	0.8761	[0.0, 0.8761025846276997]	[nan, 0.8761025846276997]
0.0117	2.73	240	0.0339	0.4471	0.8941	0.8941	[0.0, 0.8941320836457919]	[nan, 0.8941320836457919]
0.0198	2.95	260	0.0297	0.4485	0.8969	0.8969	[0.0, 0.8969491585060927]	[nan, 0.8969491585060927]
0.0247	3.18	280	0.0303	0.4565	0.9130	0.9130	[0.0, 0.9130423308930413]	[nan, 0.9130423308930413]
0.0115	3.41	300	0.0307	0.4533	0.9066	0.9066	[0.0, 0.9065626188900153]	[nan, 0.9065626188900153]
0.0164	3.64	320	0.0330	0.4549	0.9097	0.9097	[0.0, 0.9097436483868343]	[nan, 0.9097436483868343]
0.0114	3.86	340	0.0362	0.4425	0.8850	0.8850	[0.0, 0.8849727418868903]	[nan, 0.8849727418868903]
0.012	4.09	360	0.0321	0.4582	0.9164	0.9164	[0.0, 0.9164498699219532]	[nan, 0.9164498699219532]
0.0153	4.32	380	0.0321	0.4572	0.9144	0.9144	[0.0, 0.9144310762281544]	[nan, 0.9144310762281544]
0.0115	4.55	400	0.0307	0.4573	0.9145	0.9145	[0.0, 0.9145300367033407]	[nan, 0.9145300367033407]
0.0139	4.77	420	0.0330	0.4678	0.9357	0.9357	[0.0, 0.935664695520609]	[nan, 0.935664695520609]
0.014	5.0	440	0.0317	0.4635	0.9271	0.9271	[0.0, 0.9270562337402442]	[nan, 0.9270562337402442]
0.0197	5.23	460	0.0320	0.4678	0.9356	0.9356	[0.0, 0.9355745315321061]	[nan, 0.9355745315321061]
0.0086	5.45	480	0.0337	0.4607	0.9214	0.9214	[0.0, 0.9213528116870122]	[nan, 0.9213528116870122]
0.3103	5.68	500	0.0338	0.4548	0.9096	0.9096	[0.0, 0.9095853116265363]	[nan, 0.9095853116265363]
0.0088	5.91	520	0.0305	0.4635	0.9270	0.9270	[0.0, 0.9270243464760175]	[nan, 0.9270243464760175]
0.0119	6.14	540	0.0299	0.4680	0.9359	0.9359	[0.0, 0.9359494817769782]	[nan, 0.9359494817769782]
0.0114	6.36	560	0.0314	0.4574	0.9148	0.9148	[0.0, 0.914796130425508]	[nan, 0.914796130425508]
0.0122	6.59	580	0.0289	0.4613	0.9227	0.9227	[0.0, 0.9226920767845322]	[nan, 0.9226920767845322]
0.0164	6.82	600	0.0312	0.4620	0.9240	0.9240	[0.0, 0.9239807620836238]	[nan, 0.9239807620836238]
0.0062	7.05	620	0.0335	0.4605	0.9210	0.9210	[0.0, 0.9209954544155065]	[nan, 0.9209954544155065]
0.0089	7.27	640	0.0309	0.4659	0.9317	0.9317	[0.0, 0.9317029778306545]	[nan, 0.9317029778306545]
0.0251	7.5	660	0.0291	0.4734	0.9468	0.9468	[0.0, 0.9467878529315391]	[nan, 0.9467878529315391]
0.0065	7.73	680	0.0326	0.4598	0.9195	0.9195	[0.0, 0.9195297398219151]	[nan, 0.9195297398219151]
0.0056	7.95	700	0.0310	0.4606	0.9213	0.9213	[0.0, 0.9212714441851925]	[nan, 0.9212714441851925]
0.0099	8.18	720	0.0345	0.4503	0.9006	0.9006	[0.0, 0.9006183930138303]	[nan, 0.9006183930138303]
0.0103	8.41	740	0.0335	0.4539	0.9078	0.9078	[0.0, 0.9077512441530853]	[nan, 0.9077512441530853]
0.0065	8.64	760	0.0334	0.4544	0.9088	0.9088	[0.0, 0.9087936278250467]	[nan, 0.9087936278250467]
0.0047	8.86	780	0.0341	0.4557	0.9114	0.9114	[0.0, 0.9114215782216583]	[nan, 0.9114215782216583]
0.0105	9.09	800	0.0315	0.4597	0.9195	0.9195	[0.0, 0.9194703635368034]	[nan, 0.9194703635368034]
0.0087	9.32	820	0.0329	0.4583	0.9166	0.9166	[0.0, 0.9165708216138474]	[nan, 0.9165708216138474]
0.0122	9.55	840	0.0357	0.4537	0.9073	0.9073	[0.0, 0.9073004242105703]	[nan, 0.9073004242105703]
0.0057	9.77	860	0.0319	0.4621	0.9241	0.9241	[0.0, 0.9241050124580242]	[nan, 0.9241050124580242]
0.0068	10.0	880	0.0342	0.4539	0.9078	0.9078	[0.0, 0.907799624829843]	[nan, 0.907799624829843]
0.0095	10.23	900	0.0340	0.4578	0.9156	0.9156	[0.0, 0.9155933120311748]	[nan, 0.9155933120311748]
0.0043	10.45	920	0.0319	0.4636	0.9272	0.9272	[0.0, 0.9271771854321385]	[nan, 0.9271771854321385]
0.0049	10.68	940	0.0308	0.4659	0.9319	0.9319	[0.0, 0.9318525181042692]	[nan, 0.9318525181042692]
0.005	10.91	960	0.0319	0.4640	0.9281	0.9281	[0.0, 0.9280612323438019]	[nan, 0.9280612323438019]
0.0043	11.14	980	0.0313	0.4653	0.9306	0.9306	[0.0, 0.930638602941985]	[nan, 0.930638602941985]
0.0084	11.36	1000	0.0321	0.4632	0.9264	0.9264	[0.0, 0.9264294840640648]	[nan, 0.9264294840640648]
0.0044	11.59	1020	0.0320	0.4643	0.9285	0.9285	[0.0, 0.9285241474555063]	[nan, 0.9285241474555063]
0.0044	11.82	1040	0.0321	0.4661	0.9321	0.9321	[0.0, 0.9321098153397533]	[nan, 0.9321098153397533]
0.0057	12.05	1060	0.0338	0.4626	0.9253	0.9253	[0.0, 0.9252518544093489]	[nan, 0.9252518544093489]
0.0064	12.27	1080	0.0348	0.4616	0.9231	0.9231	[0.0, 0.9231450958487181]	[nan, 0.9231450958487181]
0.0075	12.5	1100	0.0331	0.4618	0.9237	0.9237	[0.0, 0.9236706859280404]	[nan, 0.9236706859280404]
0.0103	12.73	1120	0.0317	0.4704	0.9408	0.9408	[0.0, 0.9408425274945187]	[nan, 0.9408425274945187]
0.0053	12.95	1140	0.0320	0.4704	0.9407	0.9407	[0.0, 0.9407292727284723]	[nan, 0.9407292727284723]
0.0073	13.18	1160	0.0331	0.4652	0.9305	0.9305	[0.0, 0.9304681710124976]	[nan, 0.9304681710124976]
0.0052	13.41	1180	0.0342	0.4664	0.9328	0.9328	[0.0, 0.9328047377877275]	[nan, 0.9328047377877275]
0.0089	13.64	1200	0.0322	0.4676	0.9353	0.9353	[0.0, 0.9352996413232555]	[nan, 0.9352996413232555]
0.0054	13.86	1220	0.0332	0.4655	0.9311	0.9311	[0.0, 0.9310509382552609]	[nan, 0.9310509382552609]
0.0057	14.09	1240	0.0333	0.4661	0.9321	0.9321	[0.0, 0.9321439017256508]	[nan, 0.9321439017256508]
0.0047	14.32	1260	0.0346	0.4639	0.9278	0.9278	[0.0, 0.9277522557490538]	[nan, 0.9277522557490538]
0.0092	14.55	1280	0.0380	0.4583	0.9166	0.9166	[0.0, 0.9166290983381238]	[nan, 0.9166290983381238]
0.0066	14.77	1300	0.0338	0.4638	0.9277	0.9277	[0.0, 0.927687381659765]	[nan, 0.927687381659765]
0.0076	15.0	1320	0.0347	0.4640	0.9280	0.9280	[0.0, 0.9279897608895007]	[nan, 0.9279897608895007]
0.0054	15.23	1340	0.0345	0.4647	0.9295	0.9295	[0.0, 0.9294664710914461]	[nan, 0.9294664710914461]
0.0036	15.45	1360	0.0349	0.4666	0.9332	0.9332	[0.0, 0.9331950818842955]	[nan, 0.9331950818842955]
0.004	15.68	1380	0.0352	0.4617	0.9234	0.9234	[0.0, 0.9234408777134413]	[nan, 0.9234408777134413]
0.0042	15.91	1400	0.0357	0.4622	0.9244	0.9244	[0.0, 0.9244282833436326]	[nan, 0.9244282833436326]
0.0048	16.14	1420	0.0370	0.4586	0.9172	0.9172	[0.0, 0.9171546884174461]	[nan, 0.9171546884174461]
0.0043	16.36	1440	0.0345	0.4647	0.9294	0.9294	[0.0, 0.9294411811922318]	[nan, 0.9294411811922318]
0.0027	16.59	1460	0.0354	0.4667	0.9334	0.9334	[0.0, 0.9333754098613014]	[nan, 0.9333754098613014]
0.0057	16.82	1480	0.0364	0.4689	0.9379	0.9379	[0.0, 0.9378913062122988]	[nan, 0.9378913062122988]
0.0035	17.05	1500	0.0363	0.4662	0.9325	0.9325	[0.0, 0.9324682721720945]	[nan, 0.9324682721720945]
0.0029	17.27	1520	0.0348	0.4674	0.9347	0.9347	[0.0, 0.9347212723238338]	[nan, 0.9347212723238338]
0.0043	17.5	1540	0.0362	0.4648	0.9295	0.9295	[0.0, 0.9295390421065827]	[nan, 0.9295390421065827]
0.0041	17.73	1560	0.0347	0.4664	0.9328	0.9328	[0.0, 0.9328487202211436]	[nan, 0.9328487202211436]
0.003	17.95	1580	0.0364	0.4649	0.9297	0.9297	[0.0, 0.9297237683269303]	[nan, 0.9297237683269303]
0.0121	18.18	1600	0.0364	0.4650	0.9300	0.9300	[0.0, 0.9299920611707684]	[nan, 0.9299920611707684]
0.004	18.41	1620	0.0369	0.4667	0.9334	0.9334	[0.0, 0.9334259896597299]	[nan, 0.9334259896597299]
0.0035	18.64	1640	0.0368	0.4636	0.9272	0.9272	[0.0, 0.9272475573256042]	[nan, 0.9272475573256042]
0.0031	18.86	1660	0.0358	0.4665	0.9330	0.9330	[0.0, 0.9329784683997212]	[nan, 0.9329784683997212]
0.0032	19.09	1680	0.0357	0.4661	0.9322	0.9322	[0.0, 0.9321515986514985]	[nan, 0.9321515986514985]
0.0047	19.32	1700	0.0371	0.4621	0.9243	0.9243	[0.0, 0.9242886391175364]	[nan, 0.9242886391175364]
0.0056	19.55	1720	0.0359	0.4663	0.9326	0.9326	[0.0, 0.9326277084932278]	[nan, 0.9326277084932278]
0.0033	19.77	1740	0.0348	0.4694	0.9389	0.9389	[0.0, 0.9388523223824404]	[nan, 0.9388523223824404]
0.0049	20.0	1760	0.0394	0.4612	0.9224	0.9224	[0.0, 0.9223918966764674]	[nan, 0.9223918966764674]
0.0058	20.23	1780	0.0368	0.4660	0.9321	0.9321	[0.0, 0.9320724302713497]	[nan, 0.9320724302713497]
0.003	20.45	1800	0.0370	0.4686	0.9372	0.9372	[0.0, 0.9371787907909581]	[nan, 0.9371787907909581]
0.0058	20.68	1820	0.0363	0.4665	0.9330	0.9330	[0.0, 0.9329949618122522]	[nan, 0.9329949618122522]
0.0083	20.91	1840	0.0351	0.4661	0.9322	0.9322	[0.0, 0.9321834859157253]	[nan, 0.9321834859157253]
0.0036	21.14	1860	0.0353	0.4667	0.9333	0.9333	[0.0, 0.9333149340153543]	[nan, 0.9333149340153543]
0.0032	21.36	1880	0.0373	0.4657	0.9314	0.9314	[0.0, 0.93137640826254]	[nan, 0.93137640826254]
0.005	21.59	1900	0.0391	0.4647	0.9294	0.9294	[0.0, 0.929370809298766]	[nan, 0.929370809298766]
0.0049	21.82	1920	0.0364	0.4701	0.9403	0.9403	[0.0, 0.9402795523467927]	[nan, 0.9402795523467927]
0.0044	22.05	1940	0.0368	0.4672	0.9343	0.9343	[0.0, 0.9343111361322288]	[nan, 0.9343111361322288]
0.0038	22.27	1960	0.0367	0.4663	0.9325	0.9325	[0.0, 0.932513354166346]	[nan, 0.932513354166346]
0.0032	22.5	1980	0.0378	0.4679	0.9358	0.9358	[0.0, 0.9358483221801213]	[nan, 0.9358483221801213]
0.0039	22.73	2000	0.0381	0.4653	0.9306	0.9306	[0.0, 0.9305517376359882]	[nan, 0.9305517376359882]
0.0032	22.95	2020	0.0385	0.4651	0.9301	0.9301	[0.0, 0.9301262075926875]	[nan, 0.9301262075926875]
0.0058	23.18	2040	0.0381	0.4654	0.9309	0.9309	[0.0, 0.9308673115957486]	[nan, 0.9308673115957486]
0.0049	23.41	2060	0.0377	0.4658	0.9316	0.9316	[0.0, 0.9316194112071639]	[nan, 0.9316194112071639]
0.0032	23.64	2080	0.0373	0.4692	0.9384	0.9384	[0.0, 0.9384256927783043]	[nan, 0.9384256927783043]
0.0056	23.86	2100	0.0390	0.4646	0.9292	0.9292	[0.0, 0.9292465589243656]	[nan, 0.9292465589243656]
0.003	24.09	2120	0.0383	0.4658	0.9317	0.9317	[0.0, 0.9316765883706047]	[nan, 0.9316765883706047]
0.0037	24.32	2140	0.0376	0.4668	0.9337	0.9337	[0.0, 0.9336755899693663]	[nan, 0.9336755899693663]
0.0025	24.55	2160	0.0390	0.4663	0.9326	0.9326	[0.0, 0.9326145137632029]	[nan, 0.9326145137632029]
0.0039	24.77	2180	0.0381	0.4688	0.9376	0.9376	[0.0, 0.937613117320942]	[nan, 0.937613117320942]
0.0031	25.0	2200	0.0395	0.4645	0.9291	0.9291	[0.0, 0.9290629322648534]	[nan, 0.9290629322648534]
0.0026	25.23	2220	0.0389	0.4668	0.9336	0.9336	[0.0, 0.9335678330074968]	[nan, 0.9335678330074968]
0.0028	25.45	2240	0.0375	0.4680	0.9359	0.9359	[0.0, 0.9359329883644473]	[nan, 0.9359329883644473]
0.0039	25.68	2260	0.0404	0.4656	0.9312	0.9312	[0.0, 0.9312004785288756]	[nan, 0.9312004785288756]
0.004	25.91	2280	0.0371	0.4716	0.9431	0.9431	[0.0, 0.9431021250112706]	[nan, 0.9431021250112706]
0.0048	26.14	2300	0.0373	0.4700	0.9400	0.9400	[0.0, 0.9399639783870323]	[nan, 0.9399639783870323]
0.0033	26.36	2320	0.0385	0.4688	0.9377	0.9377	[0.0, 0.9376560001935227]	[nan, 0.9376560001935227]
0.0042	26.59	2340	0.0374	0.4686	0.9372	0.9372	[0.0, 0.9371743925476165]	[nan, 0.9371743925476165]
0.0048	26.82	2360	0.0393	0.4660	0.9320	0.9320	[0.0, 0.9319789676003404]	[nan, 0.9319789676003404]
0.0047	27.05	2380	0.0393	0.4650	0.9300	0.9300	[0.0, 0.9300162515091472]	[nan, 0.9300162515091472]
0.0048	27.27	2400	0.0389	0.4670	0.9340	0.9340	[0.0, 0.9339867656857851]	[nan, 0.9339867656857851]
0.004	27.5	2420	0.0388	0.4673	0.9344	0.9344	[0.0, 0.9343870058298716]	[nan, 0.9343870058298716]