SegFormer-B0 Open-Source Image Segmentation Model - Free Deployment for Semantic Segmentation of Sidewalk Scenarios

Segformer B0 Finetuned Segments Sidewalk

Developed by nielsr

An image segmentation model fine-tuned on the segments/sidewalk-semantic dataset based on nvidia/mit-b0, designed for semantic segmentation of sidewalk scenes

Image Segmentation

Transformers

Open Source License:Apache-2.0 #Sidewalk semantic segmentation #Lightweight model #Urban street scene analysis

Downloads 15

Release Time : 3/5/2022

Model Overview

This model is a lightweight version (b0) of the SegFormer architecture, specifically optimized for semantic segmentation tasks in sidewalk scenes. It can identify and segment different objects and areas in sidewalk environments.

Model Features

Lightweight design

Based on the SegFormer-b0 architecture, suitable for deployment in resource-constrained environments

Optimized for sidewalk scenes

Specifically fine-tuned for sidewalk environments to improve recognition accuracy of relevant objects

Multi-category segmentation

Capable of identifying and segmenting multiple objects and areas in sidewalk environments

Model Capabilities

Image semantic segmentation

Scene understanding

Object recognition

Use Cases

Smart cities

Sidewalk condition monitoring

Automatically identifies and segments various elements on sidewalks for urban infrastructure maintenance

Can recognize multiple elements such as roads and obstacles

Autonomous driving

Pedestrian area recognition

Assists autonomous driving systems in identifying passable sidewalk areas

Achieves an IoU of 0.717 for sidewalk recognition

🚀 segformer-b0-finetuned-segments-sidewalk

This model is a fine - tuned version of nvidia/mit-b0 on the segments/sidewalk - semantic dataset. It offers high - quality image segmentation capabilities, achieving excellent results on the evaluation set.

🚀 Quick Start

This section provides a general overview of the model. For specific usage steps, you may need to refer to the official documentation of the underlying framework.

📚 Documentation

Model Performance

This model achieves the following results on the evaluation set:

Loss: 0.5679
Miou: 0.2769
Macc: 0.3331
Overall Accuracy: 0.8424
Per Category Iou: [nan, 0.7174911859423314, 0.8790751054409742, 0.6065232798410057, 0.6975274018055722, 0.3486407385349508, nan, 0.40093167116703843, 0.28779837903852556, 0.0, 0.7870339041746186, 0.0, 0.0, 0.0, 0.0, 0.1464360606454247, 0.0, 0.0, 0.6770283275082656, 0.0, 0.338555175257431, 0.14697310016578427, 0.0, nan, 0.0, 0.27163002251763635, 0.0, 0.0, 0.8257437911843676, 0.7169333376341568, 0.9108105550493353, 0.0, 0.0, 0.1016801552778885, 0.0]
Per Category Accuracy: [nan, 0.9199960254104915, 0.9327745517652714, 0.7304629327758765, 0.7378309547498484, 0.45295941407150275, nan, 0.5188608021128075, 0.5327441812670195, 0.0, 0.9353764765979435, 0.0, 0.0, 0.0, 0.0, 0.1588525415198792, 0.0, 0.0, 0.9238854794385364, 0.0, 0.4400394213522207, 0.15130051149615126, 0.0, nan, 0.0, 0.3570096986572905, 0.0, 0.0, 0.9359897980968498, 0.8570458108260572, 0.9549583230619891, 0.0, 0.0, 0.11786971668879294, 0.0]

Training Procedure

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: 5

Training Results

Training Loss	Epoch	Step	Validation Loss	Miou	Macc	Overall Accuracy	Per Category Iou	Per Category Accuracy
1.357	1.0	400	1.0006	0.1632	0.2069	0.7524	[nan, 0.5642795884663824, 0.7491853309192827, 0.0, 0.40589649630192104, 0.02723606910696284, nan, 0.0002207740938439576, 0.0, 0.0, 0.6632462867093903, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.5671699281129761, 0.0, 0.0009207911027492868, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.7507253434892517, 0.6157793573905029, 0.8774768871968204, 0.0, 0.0, 0.0, 0.0]	[nan, 0.6839993330882016, 0.9786792586618772, 0.0, 0.4818162160949784, 0.02785198456498826, nan, 0.00022133459131411787, 0.0, 0.0, 0.9043689536433023, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.8606078323791991, 0.0, 0.0009210330367246509, 0.0, 0.0, nan, 0.0, 0.0, 0.0, 0.0, 0.895198618615298, 0.8549807032886052, 0.9328734839751688, 0.0, 0.0, 0.0, 0.0]
1.6346	2.0	800	0.7856	0.1903	0.2334	0.7917	[nan, 0.6276046255936906, 0.8379492348238635, 0.0, 0.5220035981992285, 0.19441920935217594, nan, 0.16135703555333, 0.0, 0.0, 0.7357165628674137, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.567598980063164, 0.0, 0.07867871139133086, 0.0, 0.0, nan, 0.0, 0.02123705398363847, 0.0, 0.0, 0.7917172051343153, 0.6589515948064048, 0.8916684207946344, 0.0, 0.0, 0.00013685918191589503, 0.0]	[nan, 0.8610263337355926, 0.9499345560017969, 0.0, 0.5908796687797819, 0.2144081438468206, nan, 0.1813236746419022, 0.0, 0.0, 0.8825551027577866, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.9239907140298015, 0.0, 0.08495225520298297, 0.0, 0.0, nan, 0.0, 0.021302829364985724, 0.0, 0.0, 0.9258397010509258, 0.8834861376443207, 0.9489131468773239, 0.0, 0.0, 0.0001372777815910495, 0.0]
0.659	3.0	1200	0.6798	0.2215	0.2687	0.8107	[nan, 0.6728474586764454, 0.8404607924530816, 0.21147709475332813, 0.5407350347311378, 0.23535489130104167, nan, 0.3087159264982809, 0.0060319580742948155, 0.0, 0.7331305064022374, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.6378031991744924, 0.0, 0.35289337122777764, 6.24997656258789e - 05, 0.0, nan, 0.0, 0.14698390926256938, 0.0, 0.0, 0.8019042204623998, 0.669283249725758, 0.8928145424856038, 0.0, 0.0, 0.03847722460691187, 0.0]	[nan, 0.866012011452706, 0.9627112260298595, 0.21236715482371135, 0.5645869262075475, 0.2750610095322395, nan, 0.3857655597748765, 0.0060319580742948155, 0.0, 0.939196440844118, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.8380282443529743, 0.0, 0.5749902063170915, 6.256068386334744e - 05, 0.0, nan, 0.0, 0.1605725590139305, 0.0, 0.0, 0.9212803460870584, 0.8870298583701837, 0.959700359744241, 0.0, 0.0, 0.04453994364914478, 0.0]
0.5481	4.0	1600	0.5999	0.2522	0.2998	0.8312	[nan, 0.7078353465279917, 0.8661728761172196, 0.3857324719136883, 0.6338278880825696, 0.3440050078187208, nan, 0.35980405625532347, 0.23875867241702606, 0.0, 0.773703347865372, 0.0, 0.0, 0.0, 0.0, 0.0004931363471679884, 0.0, 0.0, 0.6554146448850521, 0.0, 0.367673493717809, 0.03089804641909161, 0.0, nan, 0.0, 0.21529017459808872, 0.0, 0.0, 0.818951849158376, 0.7007504838794707, 0.9053929635423027, 0.0, 0.0, 0.06626212301200333, 0.0]	[nan, 0.8955207784307155, 0.9536263694097721, 0.39712577675621036, 0.6989299616008556, 0.4248959179453637, nan, 0.42984959564233455, 0.26168627652468784, 0.0, 0.9055166364779607, 0.0, 0.0, 0.0, 0.0, 0.0004932058379466533, 0.0, 0.0, 0.8632164276000204, 0.0, 0.6365580872107307, 0.031401709658368616, 0.0, nan, 0.0, 0.2497286263775161, 0.0, 0.0, 0.9296676429517725, 0.8858954297713482, 0.9555756265860916, 0.0, 0.0, 0.0750792276952902, 0.0]
0.7855	5.0	2000	0.5679	0.2769	0.3331	0.8424	[nan, 0.7174911859423314, 0.8790751054409742, 0.6065232798410057, 0.6975274018055722, 0.3486407385349508, nan, 0.40093167116703843, 0.28779837903852556, 0.0, 0.7870339041746186, 0.0, 0.0, 0.0, 0.0, 0.1464360606454247, 0.0, 0.0, 0.6770283275082656, 0.0, 0.338555175257431, 0.14697310016578427, 0.0, nan, 0.0, 0.27163002251763635, 0.0, 0.0, 0.8257437911843676, 0.7169333376341568, 0.9108105550493353, 0.0, 0.0, 0.1016801552778885, 0.0]	[nan, 0.9199960254104915, 0.9327745517652714, 0.7304629327758765, 0.7378309547498484, 0.45295941407150275, nan, 0.5188608021128075, 0.5327441812670195, 0.0, 0.9353764765979435, 0.0, 0.0, 0.0, 0.0, 0.1588525415198792, 0.0, 0.0, 0.9238854794385364, 0.0, 0.4400394213522207, 0.15130051149615126, 0.0, nan, 0.0, 0.3570096986572905, 0.0, 0.0, 0.9359897980968498, 0.8570458108260572, 0.9549583230619891, 0.0, 0.0, 0.11786971668879294, 0.0]

Framework Versions

Transformers 4.17.0
Pytorch 1.10.0+cu111
Datasets 1.18.3
Tokenizers 0.11.6

📄 License

This model is licensed under the Apache - 2.0 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