Segformer B0开源图像分割模型 - 免费部署，专为场景解析任务优化！

首页

Segformer B0 Scene Parse 150

由 univers1123 开发

基于MIT-B0架构的轻量级图像分割模型，专为场景解析任务优化

图像分割

Transformers

开源协议:其他 #场景解析 #图像语义分割 #MIT-B0架构

下载量 20

发布时间 : 8/7/2023

模型简介

该模型是基于NVIDIA的MIT-B0架构在Scene Parse 150数据集上微调的场景解析模型，能够对复杂场景中的150类物体进行像素级分割识别

模型特点

轻量级架构

采用MIT-B0轻量级主干网络，平衡性能与计算效率

多类别识别

支持150类场景物体的精确分割识别

端到端训练

完整训练流程优化，可直接应用于实际场景解析任务

模型能力

图像语义分割

场景理解

像素级分类

多物体识别

使用案例

智能城市

街景分析

自动识别街道场景中的各类元素（建筑、道路、车辆等）

自动驾驶

环境感知

实时解析道路场景，辅助车辆决策系统

🚀 segformer-b0-scene-parse-150

本模型是在scene_parse_150数据集上对nvidia/mit-b0进行微调后的版本，可用于图像分割任务，在评估集上取得了一系列评估指标结果。

🚀 快速开始

该模型是 nvidia/mit-b0 在 scene_parse_150 数据集上的微调版本。它在评估集上取得了以下结果：

损失值：2.3431
平均交并比（Mean Iou）：0.0959
平均准确率（Mean Accuracy）：0.1537
整体准确率（Overall Accuracy）：0.5496
每类别交并比（Per Category Iou）：[0.44824978876617866, 0.7548671615728508, 0.7119201505944329, 0.5304481563680256, 0.5684691275095736, 0.33051502835188457, 0.6982393617021276, 0.0, 0.3703529914609331, 0.6659141206351092, 0.028823893043720683, 0.17181416221210322, 0.052153820762502065, 0.0, 0.0, 0.0005543923800536699, 0.40565901784724534, 0.05230759173712194, 0.0, 0.07225859019823891, 0.29980315155352005, nan, 0.003601361102652032, 0.0, 0.0, nan, 0.0, 0.0, 0.38898705304076847, 0.05940808241958817, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]
每类别准确率（Per Category Accuracy）：[0.8427949438202247, 0.9402615186644498, 0.7846678763016725, 0.7286579984703183, 0.8303175022736334, 0.469325820621132, 0.9020126572710594, nan, 0.5974398752913491, 0.9683369330453564, 0.05725843345934362, 0.24220857754209693, 0.12377594986290638, 0.0, 0.0, 0.0005611873291065182, 0.9580213623749935, 0.08566177782535773, 0.0, 0.16335928996064641, 0.43531591571750716, nan, 0.0036190907034607555, 0.0, 0.0, nan, nan, 0.0, 0.45750991876062724, 0.24276243093922653, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]

🔧 技术细节

训练超参数

训练过程中使用了以下超参数：

学习率（learning_rate）：6e - 05
训练批次大小（train_batch_size）：2
评估批次大小（eval_batch_size）：2
随机种子（seed）：42
优化器（optimizer）：Adam，其中betas=(0.9, 0.999)，epsilon=1e - 08
学习率调度器类型（lr_scheduler_type）：线性
训练轮数（num_epochs）：50

训练结果

训练损失	轮数	步数	验证损失	平均交并比	平均准确率	整体准确率	每类别交并比	每类别准确率
4.9918	0.5	20	4.8969	0.0108	0.0487	0.1875	[0.18900717264720193, 0.17829851112253592, 0.40144144917749963, 0.1885612981412077, 0.11895876927062042, 0.09866217819019046, 0.0057814729592400894, 0.0, 0.0, 0.0, 0.009622579129617706, 0.022129523898301137, 0.0037298450062015365, 0.0, 0.0, 0.0, 0.06277911646586345, 0.0, 0.0, 0.003906402593851322, 0.012887091043266734, nan, 0.0019786836291242806, 0.0, 0.0, 0.0, 0.0, 0.0, 0.015807537456273512, 0.016491354532320934, 0.0, 0.0, 0.0, nan, 0.001438298321545445, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.025794247180438844, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, nan, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.012904182735093445, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0]	[0.2067212858926342, 0.2463388525747603, 0.8113718838750394, 0.515406462102938, 0.1316758582686337, 0.11907251217424253, 0.007544887960475232, nan, 0.0, 0.0, 0.013315354795213214, 0.22085775420969392, 0.054576315445880666, 0.0, 0.0, 0.0, 0.07673176606105031, 0.0, 0.0, 0.004186552792430713, 0.013544374703761687, nan, 0.0021687933259709673, 0.0, 0.0, nan, nan, 0.0, 0.01809937653504629, 0.30082872928176796, 0.0, nan, 0.0, nan, 0.0019430975470621792, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.3248302818350134, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.05007914807886027, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]
4.551	1.0	40	4.5955	0.0202	0.0640	0.3442	[0.3519414971273263, 0.3937735618735424, 0.42161939446421154, 0.21975617697678057, 0.3809140886893701, 0.09030492572322127, 0.005777833411293457, 0.0, 0.0, 0.0, 0.0, 0.02885598249784122, 0.0, 0.0, 0.0, 0.0, 0.0573680633208358, 0.0, 0.0011308737583006134, 0.006298751950078003, 0.057306667023884476, nan, 0.0014234124996705063, 0.0, 0.0, 0.0, nan, 0.0, 0.017088433502956954, 0.023128390596745027, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0052120890103356235, 0.0, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, 0.0, 0.0, nan, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, 0.0]	[0.5003401997503121, 0.785197975265903, 0.952571789207952, 0.4632600048849975, 0.5973921536125012, 0.09887449654069073, 0.006026290292656446, nan, 0.0, 0.0, 0.0, 0.4399013861754582, 0.0, 0.0, 0.0, 0.0, 0.06849835069898948, 0.0, 0.0098046905639658, 0.0067612827597756005, 0.06144521207072375, nan, 0.0014369918969623589, 0.0, 0.0, nan, nan, 0.0, 0.018779520120914415, 0.07066298342541437, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.01695124459987657, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]
4.3293	1.5	60	4.2491	0.0352	0.0776	0.4184	[0.3871309742960387, 0.45006666311952553, 0.6112315905191344, 0.3032571607536305, 0.44533070206501846, 0.063346836376098, 0.022528980712202534, 0.0, 0.0, 0.0, 0.0, 0.029790687595074403, 0.0019532612486920127, 0.0, 0.0, 0.0, 0.24116048081196448, 0.0, 0.0, 0.0003640114056907116, 0.2558965972702686, nan, 0.0005833454863642993, 0.0, 0.0, nan, 0.0, 0.0, 0.002751498247333308, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]	[0.7088701622971286, 0.8308114309791149, 0.8838434837488167, 0.5903983055833383, 0.8704499330230141, 0.06874412338436925, 0.02526644174013427, nan, 0.0, 0.0, 0.0, 0.3261233603125872, 0.0036558297427862646, 0.0, 0.0, 0.0, 0.305421749829834, 0.0, 0.0, 0.00037678975131876413, 0.350616893842552, nan, 0.0005854411432068869, 0.0, 0.0, nan, nan, 0.0, 0.0027583600982429624, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]
3.9435	2.0	80	4.1371	0.0349	0.0760	0.4181	[0.35782571228523724, 0.4339785275779836, 0.33734326770427064, 0.3463785302488506, 0.485148026657255, 0.060373176138162864, 0.25141442893216265, 0.0, 0.0, 0.0, 0.0, 0.03557946863062567, 0.0016987542468856172, 0.0, 0.0, 0.0, 0.07208336234305739, 0.0, 0.0, 0.0, 0.23636984914447706, nan, 0.0012355355980390855, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0013693940431359123, 0.0, nan, 0.0, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]	[0.5806023720349563, 0.8385180609014508, 0.9781397917324077, 0.6769715154724274, 0.8375259283816657, 0.06956509535905245, 0.3441307230861203, nan, 0.0, 0.0, 0.0, 0.18348218438924552, 0.0019584802193497847, 0.0, 0.0, 0.0, 0.08131315775695062, 0.0, 0.0, 0.0, 0.26825905232466285, nan, 0.0012374096890509201, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0026519337016574587, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]
3.951	2.5	100	3.7658	0.0401	0.0786	0.4477	[0.405717317146636, 0.4086408628919567, 0.46930568969097103, 0.3717459827442089, 0.4797755033427323, 0.09497839561115337, 0.3326272962505519, 0.0, 0.0, 0.0, 0.0, 0.042534003647694614, 0.0014730878186968838, 0.0, 0.0, 0.0, 0.0077221054763652935, 0.0, 0.0, 0.0, 0.1928527248986324, nan, 0.0, 0.0, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]	[0.6074812734082397, 0.8766682989045027, 0.9717970968759861, 0.7857572013599547, 0.917421184439835, 0.10295479742341103, 0.40897690494678035, nan, 0.0, 0.0, 0.0, 0.14699041771327565, 0.00169734952343648, 0.0, 0.0, 0.0, 0.007984711241426252, 0.0, 0.0, 0.0, 0.2015282306134467, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, nan, nan, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, 0.0, 0.0, nan, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, 0.0, nan, nan, nan, nan, 0.0, 0.0, 0.0, nan, nan, nan, nan, nan, nan, nan, 0.0, nan, nan, 0.0, nan, nan, 0.0, 0.0, nan, nan, nan, nan, 0.0, nan, 0.0]
3.7065	3.0	120	3.6117	0.0484	0.0779	0.4436	[0.30003355195190606, 0.6100469285116588, 0.7166468505013384, 0.2808355790192315, 0.4914731888407471, 0.004109398150913026, 0.3171473942892336, 0.0, 0.0, 0.0, 0.0, 0.0061059528193994845, 0.0, 0.0, 0.0, 0.0, 0.29060069752832357, 0.0, 0.0, 0.0, 0.17904034375746, nan, 0.0, 0.0, 0.0, nan, nan, 0.0, 0.0, 0.0, 0.0, nan, 0.0, nan, 0.0, nan, 0.0, 0.0, nan, 0.0, 0.0, ...（表格后续内容过长，此处省略）	...（表格后续内容过长，此处省略）