microsoft/usroaddetections

Road detections from aerial imagery in the US

Introduction

This data is freely available for download and use and contains

1. 54,484,737 computer generated roads in all US states except Alaska.
2. 5,931,242 computer generated roads in all US states except Alaska that are missing in OpenStreetMaps roads drop from 02-May-2020

License

This data is licensed by Microsoft under the Open Data Commons Open Database License (ODbL)

FAQ

What the data include:

1. 54,484,737 computer generated roads in all US states except Alaska.
2. 5,931,242 computer generated roads in all US states except Alaska that are missing in OpenStreetMaps (OSM) roads drop from 02-May-2020

What is the GeoJson format?

GeoJSON is a format for encoding a variety of geographic data structures. For Intensive Documentation and Tutorials, Refer to GeoJson Blog

Creation Details:

The road extraction is done in four stages (first dataset went through two stages and second went through all four):

1. Semantic Segmentation – Recognizing road pixels on the aerial image using Convolutional Neural Network (CNN).
2. Geometry Generation - A series of algorithms and processes transforming output of semantic segmentation into roads in geometry format.
3. Conflation & Cutting - Excluding roads and parts of roads that already exist in the road network (OSM).
4. Classification - A classifier to filter out low-confidence roads and predict a road type.

Scheme

CNN architecture

Our network was based on UNet and ResNet and the following papers [U-Net] (https://arxiv.org/abs/1505.04597), [Res U-Net] (https://arxiv.org/pdf/1512.03385.pdf), [Res U-Net] (https://arxiv.org/pdf/1711.10684.pdf). The model was trained on 512x512 images, it is fully-convolutional, meaning that the model can be applied to an image of any size (constrained by GPU memory, 1088x1088 in our case).

Training details

The training set consists of 16800 labeled images. Majority of the satellite images cover diverse residential areas in the US. For the sake of good set representation, we have enriched the set with samples from various areas covering mountains, glaciers, forests, deserts, beaches, coasts, etc. Images in the set are of 1024x1024 pixel size with 1 meter/pixel resolution. The training is done with Keras toolkit.

Metrics

These are the intermediate stage metrics we use to track CNN model improvements and they are pixel based.
Pixel precision/recall = 83.06%/80.74%

Description

Geometry generation consists of the following steps

1. Image postprocessing
2. Thinning
3. Connectivity improvement
4. Graph construction
5. Finalizing road shapes and network quality
6. Stiching road geojsons between neighboring images where needed

Metrics

We use APLS metric to evaluate connectivity. It is measured over images with scale 200x200 meters.
APLS precision/recall = 77.61%/71.52%

Data Vintage

The vintage of the roads depends on the vintage of the underlying imagery. Because Bing Imagery is a composite of multiple sources it is difficult to know the exact dates for individual pieces of data.

How good are the data?

The Osm Missing Data went through a final classifier to ensure that the precision is at least 90%. Here is another measurement with human OSM editors before the final classifier:

Label	%
Roads added without editing	77%
Roads added with minor editing	18%
Incorrect roads	5%

Will there be more data coming for other geographies?

Yes, we are working on adding more countries. Next targets are South America and Europe

Why is the data being released?

Microsoft has a continued interest in supporting a thriving OpenStreetMap ecosystem.

Should we import the data into OpenStreetMap?

This dataset was shared with Facebook, owner or RapID - a tool for adding mined roads to OSM.

External References

Full drop of all USA mined roads				OSM missing roads (USA 02-May-2020)
State	Number of Roads	Length km	Unzipped MB	State	Number of Roads	Length km	Unzipped MB
USA	54484737	9308940	13459	USA	5931242	817761	2924
Alabama	1030015	196256	268	Alabama	169926	22311	83
Arizona	1180273	160617	263	Arizona	66934	10377	32
Arkansas	767321	174286	208	Arkansas	154038	26759	76
California	4009971	470500	889	California	252553	29645	121
Colorado	1082238	181699	268	Colorado	76741	11584	38
Connecticut	427923	51203	103	Connecticut	38427	3154	18
Delaware	141175	17592	32	Delaware	10225	969	4
Florida	2712701	346693	577	Florida	147038	19932	69
Georgia	1749917	283913	437	Georgia	191975	24302	94
Hawaii	137705	14953	29	Hawaii	13962	1334	6
Idaho	562235	114038	162	Idaho	73548	11416	37
Illinois	1840754	309918	407	Illinois	174399	21305	83
Indiana	1206595	213000	276	Indiana	146195	16725	69
Iowa	820387	222015	203	Iowa	94436	11551	44
Kansas	885228	253328	219	Kansas	93413	13423	43
Kentucky	965199	171887	269	Kentucky	190964	24101	96
Louisiana	761323	150911	187	Louisiana	127622	20813	61
Maine	464244	89907	140	Maine	80845	12431	41
Maryland	800548	88361	183	Maryland	39237	3399	18
Massachusetts	758191	83294	178	Massachusetts	34552	2758	16
Michigan	1712266	279825	404	Michigan	177966	19051	85
Minnesota	1393249	296632	355	Minnesota	191433	26747	94
Mississippi	671114	156392	185	Mississippi	134654	22403	66
Missouri	1300357	281199	351	Missouri	153014	21355	76
Montana	860457	180023	249	Montana	130533	24636	68
Nebraska	663577	187192	171	Nebraska	47586	7036	23
Nevada	463933	70773	107	Nevada	28404	4568	13
NewHampshire	243910	38689	68	NewHampshire	21741	2122	10
NewJersey	942622	98892	210	NewJersey	65934	5681	31
NewMexico	680561	127638	168	NewMexico	60150	10509	29
NewYork	1584215	251629	404	NewYork	177720	18849	87
NorthCarolina	1769687	288948	461	NorthCarolina	230270	28538	113
NorthDakota	581985	164082	154	NorthDakota	33947	4780	16
Ohio	1795292	277164	420	Ohio	193883	21635	94
Oklahoma	1014493	239617	250	Oklahoma	129390	21496	63
Oregon	1111082	173778	302	Oregon	129581	20241	67
Pennsylvania	1922526	286625	492	Pennsylvania	223364	27456	111
RhodeIsland	126379	12978	28	RhodeIsland	7863	610	3
SouthCarolina	884137	153987	228	SouthCarolina	102010	13842	50
SouthDakota	530158	151645	140	SouthDakota	69851	11593	34
Tennessee	1231633	212032	333	Tennessee	240815	34606	122
Texas	4901657	831644	1146	Texas	521434	88391	253
Utah	549837	86304	133	Utah	45061	7241	22
Vermont	164249	32170	50	Vermont	30386	3527	15
Virginia	1466999	208827	378	Virginia	108239	11788	54
Washington	1565086	209696	390	Washington	158898	21397	80
WestVirginia	446393	85916	140	WestVirginia	102378	15400	53
Wisconsin	1231694	241375	309	Wisconsin	172143	20430	83
Wyoming	371246	88897	109	Wyoming	65564	13543	33

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

Legal Notices

Microsoft, Windows, Microsoft Azure and/or other Microsoft products and services referenced in the documentation may be either trademarks or registered trademarks of Microsoft in the United States and/or other countries. The licenses for this project do not grant you rights to use any Microsoft names, logos, or trademarks. Microsoft's general trademark guidelines can be found at http://go.microsoft.com/fwlink/?LinkID=254653.

Privacy information can be found at https://privacy.microsoft.com/en-us/

Microsoft and any contributors reserve all other rights, whether under their respective copyrights, patents, or trademarks, whether by implication, estoppel or otherwise.

usroaddetections's Issues

This repo is missing a LICENSE file

This repository is currently missing a LICENSE file.

A license helps users understand how to use your project in a compliant manner. You can find the standard MIT license Microsoft uses at: https://github.com/microsoft/repo-templates/blob/main/shared/LICENSE.

If you would like to learn more about open source licenses, please visit the document at https://aka.ms/license (Microsoft-internal guidance).