Dataset Card for panoramic street view images (v.0.0.2)

  Dataset Summary

The random streetview images dataset are labeled, panoramic images scraped from randomstreetview.com. Each image shows a location accessible by Google Streetview that has been roughly combined to provide ~360 degree view of a single location. The dataset was designed with the intent to geolocate an image purely based on its visual content.

  Supported Tasks and Leaderboards

None as of now!… See the full description on the dataset page: https://huggingface.co/datasets/stochastic/random_streetview_images_pano_v0.0.2.

Streetview by Country

Intro

Hi, I thought I would pull together a dataset comprising of streetview images from different countries. I saw there were a couple datasets on Kaggle with a similar intention, but I wanted to help solve 2 main problems.

1) Balanced classes across countries. 2) Images tagged with coordinates.

What is included

What I have here are roughly 1000 images (and each image is 640x640 pixels) per country/region (many island nations get separate groupings to their nominal sovereign state, e.g., Bermuda, Faroe etc.) that have sufficient official google streetview coverage. There are some countries that do have streetview coverage, but I have not included, because the coverage is extremely limited. For example, Belarus only has limited trekker coverage in the centre of Minsk, so I don't feel it is representative of the region at large in a geography-prediction dataset. More info on which countries are in the dataset and which aren't is in Country-picking.md.

There are 111 regions in total. Each is indexed by its 2-letter shortening as you can find here https://www.iban.com/country-codes, they are also in country-picking.md, for reference.

Methodology

To collect each set I first used an incredible tool to generate valid coordinates within a specified country, you can find this at https://github.com/slashP/Vali. Big thanks to the creator I could not have created this otherwise. The coordinates generated per country are intended to be spread apart, within respective sub-regions, which I hope will help create a representative dataset of that country overall. Downloaded images were then assorted to their respective country's folder and titled with the coordinates that they correspond to. For example

"42d436079_1d473612_h213.jpg"

means that the image was taken at a latitude of 42.436079, a longitude of 1.473612 and a heading (between 0 and 360) of 213.

Another note! Some regions are just that small that I couldn't curate 1000 locations within them, or I had to scale down the minimum distance between points. Notes explaining on which countries this is relevant are in Country-picking.md.

This is my first dataset so please if you have any suggestions please let me know I'd be happy to help make this a better dataset. In the future I'd love to expand this dataset into specific subregions of countries.

This dataset provides a list Latitude and longitude coordinates of valid Google Maps Street View locations. Also includes ISO code of country. Collected using public Google Street View Static API.

Dataset Card for Street View House Numbers

The Street View House Numbers (SVHN) dataset is a large real-world image dataset used for developing machine learning and object recognition algorithms. It contains over 600,000 labeled images of house numbers taken from Google Street View. The images are cropped to a fixed resolution of 32x32 pixels, centered around a single character but may contain some distractors at the sides. SVHN is similar to the MNIST dataset but incorporates… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/StreetViewHouseNumbers.

The authoritative City of Sioux Falls street map(s).

Google street view from line to point layer. Updated as necessary.

SVHN is a real-world image dataset for developing machine learning and object recognition algorithms with minimal requirement on data preprocessing and formatting. It can be seen as similar in flavor to MNIST (e.g., the images are of small cropped digits), but incorporates an order of magnitude more labeled data (over 600,000 digit images) and comes from a significantly harder, unsolved, real world problem (recognizing digits and numbers in natural scene images). SVHN is obtained from house numbers in Google Street View images.

Annotations of critical electrical infrastructure from 2,221 Google Street View (GSV) images are provided here. Wires and poles observed in the scene were annotated by drawing lines over the wires and polygons encapsulating the poles, and creating a binary mask. The collected GSV images and corresponding binary masks are provided here and are suitable for semantic segmentation research. For ease of use for different research applications, binary masks are divided into images containing power wires only, power poles only, or both.

Variety of freely available Ordnance Survey digital mapping datasets including postcodes and administration boundaries. These datasets can be useful in helping to map a number of other datasets available on the London Datastore such as Borough or Ward level data.

The following OS products are available to download from the OS OpenData website:

• MiniScale®
• 1:250 000 Scale Colour Raster
• OS Street View®
• Boundary-Line™
• Code-Point Open®
• 1:50 000 Scale Gazetteer
• Strategi®
• Meridian™ 2
• OS Locator™
• OS Terrain 50
• Land-Form PANORAMA®
• OS VectorMap® District (vector)
• OS VectorMap® District (raster)

https://www.ordnancesurvey.co.uk/business-and-government/products/finder.html?Licensed%20for=OpenData%20(Free)&withdrawn=on">Click here to visit the Ordnance Survey OpenData pages

https://www.ordnancesurvey.co.uk/opendatadownload/products.html">Click here to download the Ordnance Survey OpenData files

Paris Street View dataset is a dataset of street view images.

Detroit Street View (DSV) is an urban remote sensing program run by the Enterprise Geographic Information Systems (EGIS) Team within the Department of Innovation and Technology at the City of Detroit. The mission of Detroit Street View is ‘To continuously observe and document Detroit’s changing physical environment through remote sensing, resulting in freely available foundational data that empowers effective city operations, informed decision making, awareness, and innovation.’ LiDAR (as well as panoramic imagery) is collected using a vehicle-mounted mobile mapping system.

Due to variations in processing, index lines are not currently available for all existing LiDAR datasets, including all data collected before September 2020. Index lines represent the approximate path of the vehicle within the time extent of the given LiDAR file. The actual geographic extent of the LiDAR point cloud varies dependent on line-of-sight.

Compressed (LAZ format) point cloud files may be requested by emailing gis@detroitmi.gov with a description of the desired geographic area, any specific dates/file names, and an explanation of interest and/or intended use. Requests will be filled at the discretion and availability of the Enterprise GIS Team. Deliverable file size limitations may apply and requestors may be asked to provide their own online location or physical media for transfer.

LiDAR was collected using an uncalibrated Trimble MX2 mobile mapping system. The data is not quality controlled, and no accuracy assessment is provided or implied. Results are known to vary significantly. Users should exercise caution and conduct their own comprehensive suitability assessments before requesting and applying this data.

Sample Dataset: https://detroitmi.maps.arcgis.com/home/item.html?id=69853441d944442f9e79199b57f26fe3

Street View is a large-scale dataset of 3D scenes, consisting of millions of images taken from street-level cameras.

This statistic shows the results of a survey on the usage of the internet to look at cities and streets (e.g. via Google Street View) online in Germany from 2013 to 2015. In 2013, there were about **** million people among the German-speaking population aged 14 years and older who frequently viewed cities and streets online.

Urban Streetscape Dataset for Vision Language Models

A curated subset of 10,000 street view images with 25 essential features optimized for training vision language models on urban environment analysis tasks.

  Dataset Description

This dataset contains street view imagery paired with comprehensive annotations covering infrastructure characteristics, visual perception metrics, environmental context, and semantic segmentation data. This comprehensive dataset represents a… See the full description on the dataset page: https://huggingface.co/datasets/Sadhana-24/StreetView-Image-Dataset-10K.

The perceived wealth and physical disorder scores annotation dataset comprises 40,000 Chinese street view images. Half of these images are annotated with perceived wealth scores ranging from 0 to 10, while the other half are annotated with perceived physical disorder scores. Higher scores indicate a greater sense of affluence and disorder in the street scene. These images were annotated by Chinese urban planners. All images are in JPG format and have been uploaded as compressed files. The corresponding perception scores for each image are stored in two CSV tables. These two types of files can be matched using image IDs. Researchers can use this dataset as it is, or augment and supplement it, in order to train their own artificial intelligence perception models to perform large-scale, real-time, automatic inference on other street view images in Chinese cities. This largely overcomes the local prediction bias caused by commonly used datasets such as the MIT Place Pulse Perception Dataset, which rely on Western street view images and Western annotators.Source publication: Zhang, Y., You, Y., Chen, S., & Cai, L. (2025). Geospatial dataset on human perceptions of wealth and physical disorder in urban China using street view imagery and deep learning. Data in Brief, 112116.

The digital map market, currently valued at $25.55 billion in 2025, is experiencing robust growth, projected to expand at a Compound Annual Growth Rate (CAGR) of 13.39% from 2025 to 2033. This expansion is fueled by several key drivers. The increasing adoption of location-based services (LBS) across diverse sectors like automotive, logistics, and smart city initiatives is a primary catalyst. Furthermore, advancements in technologies such as AI, machine learning, and high-resolution satellite imagery are enabling the creation of more accurate, detailed, and feature-rich digital maps. The shift towards cloud-based deployment models offers scalability and cost-effectiveness, further accelerating market growth. While data privacy concerns and the high initial investment costs for sophisticated mapping technologies present some challenges, the overall market outlook remains overwhelmingly positive. The competitive landscape is dynamic, with established players like Google, TomTom, and ESRI vying for market share alongside innovative startups offering specialized solutions. The segmentation of the market by solution (software and services), deployment (on-premise and cloud), and industry reveals significant opportunities for growth in sectors like automotive navigation, autonomous vehicle development, and precision agriculture, where real-time, accurate mapping data is crucial. The Asia-Pacific region, driven by rapid urbanization and technological advancements in countries like China and India, is expected to witness particularly strong growth. The market's future hinges on continuous innovation. We anticipate a rise in the demand for 3D maps, real-time updates, and integration with other technologies like the Internet of Things (IoT) and augmented reality (AR). Companies are focusing on enhancing the accuracy and detail of their maps, incorporating real-time traffic data, and developing tailored solutions for specific industry needs. The increasing adoption of 5G technology promises to further boost the market by enabling faster data transmission and real-time updates crucial for applications like autonomous driving and drone delivery. The development of high-precision mapping solutions catering to specialized sectors like infrastructure management and disaster response will also fuel future growth. Ultimately, the digital map market is poised for continued expansion, driven by technological advancements and increased reliance on location-based services across a wide spectrum of industries. Recent developments include: December 2022 - The Linux Foundation has partnered with some of the biggest technology companies in the world to build interoperable and open map data in what is an apparent move t. The Overture Maps Foundation, as the new effort is called, is officially hosted by the Linux Foundation. The ultimate aim of the Overture Maps Foundation is to power new map products through openly available datasets that can be used and reused across applications and businesses, with each member throwing their data and resources into the mix., July 27, 2022 - Google declared the launch of its Street View experience in India in collaboration with Genesys International, an advanced mapping solutions company, and Tech Mahindra, a provider of digital transformation, consulting, and business re-engineering solutions and services. Google, Tech Mahindra, and Genesys International also plan to extend this to more than around 50 cities by the end of the year 2022.. Key drivers for this market are: Growth in Application for Advanced Navigation System in Automotive Industry, Surge in Demand for Geographic Information System (GIS); Increased Adoption of Connected Devices and Internet. Potential restraints include: Complexity in Integration of Traditional Maps with Modern GIS System. Notable trends are: Surge in Demand for GIS and GNSS to Influence the Adoption of Digital Map Technology.

Folder Contents Descriptioncode.zipThis zip file contains all the code files for the experiment. After extracting, you will have access to the scripts, functions, and source code necessary to run the experiment. For a detailed explanation of the code structure and functionality, please refer to the README-code.md file.data.zipThis zip file includes all the data files used in the experiment. It contains input data, experimental datasets, and other files related to the experiment. For a detailed explanation of the data and how to use it, please refer to the README-data.md file.Code overview（1）Baselines pretraining - documented in the /CaseStudy1/base_line_pretrain.ipynb and /CaseStudy2/base_line_pretrain.ipynb files（2） Baselines training and evaluation - documented in the /CaseStudy1/base_line_train.ipynb and /CaseStudy2/base_line_train.ipynb files（3） Ablation experiments pretraining - documented in the /CaseStudy1/ablation_experiment_pretrain.ipynb and /CaseStudy2/ablation_experiment_pretrain.ipynb files（4） Ablation experiments training and evaluation - documented in the /CaseStudy1/ablation_experiment_train.ipynb and /CaseStudy2/ablation_experiment_train.ipynb files（5） t-SNE Visualization for semantic knowledge between geographical locations (Base on case study 1)- documented in the /CaseStudy1/utils/Visualization/t-SNE.py file（6）GradCAM++ for street view imagery (Base on case study 1)- documented in the /CaseStudy1/utils/Visualization/cam_main.py file（7）Using SHAP to analyse the importance of viusal knowledge in case study 1- documented in the /CaseStudy1/utils/Visualization/SHAP.py file（8） Using integrated gradient to analyse the importance of viusal knowledge in case study 2- documented in the /CaseStudy2/utils/IntegratedGradient/main.py fileData OverviewThe data directory contains all the datasets and related files used in the experiment.Case study 1: Urban village classificationCaseStudy1/├── data/| ├── train_img.npy # the npy file converted from the original street view images of the training set, shape: (n_train, 4, 3, 256, 256)| ├── train_feat.npy # the visual knowledge of the training set, shape: (n_train, 128)| ├── train_svid.npy # the street view image IDs of the training set, shape: (n_train,)| ├── train_y.npy # the label of the training set, shape: (n_train,)| ├── val_img.npy # the npy file converted from the original street view images of the validation set, shape: (n_val, 4, 3, 256, 256)| ├── val_feat.npy: # the visual knowledge of the validation set, shape: (n_val, 128)| ├── val_svid.npy # the street view image IDs of the validation set, shape: (n_val,)| ├── val_y.npy # the label of the validation set, shape: (n_val,)| ├── dist.npy # the distance matrix, shape: (n, n)| ├── svids.npy # the street view image IDs corresponding to the distance matrix, shape: (n,)with: n: the number of samples n_train: the number of training samples n_val: the number of validation samples n_train : n_val = 7 : 3Case study 2: Urban mobility pattern predictionCaseStudy2/├── data/ | ├── pretrain/| ├── train_img.npy # the npy file converted from the original street view images of the training set, shape: (n_train, 3, 256, 256)| ├── train_feat.npy # the visual knowledge of the training set, shape: (n_train, 30)| ├── train_svid.npy # the street view image IDs of the training set, shape: (n_train,)| ├── val_img.npy # the npy file converted from the original street view images of the validation set, shape: (n_val, 3, 256, 256)| ├── val_feat.npy: # the visual knowledge of the validation set, shape: (n_val, 30)| ├── val_svid.npy # the street view image IDs of the validation set, shape: (n_val,)| ├── dist.npy # the distance matrix, shape: (n, n)| ├── svids.npy # the street view image IDs corresponding to the distance matrix, shape: (n,)| ├── train/| ├── feats/ | ├── 1.npy # the npy file represents the visual knowledge corresponding to the region with ID 1， shape (num_SVI_1, 30)| ├── 2.npy # the npy file represents the visual knowledge corresponding to the region with ID 2， shape (num_SVI_2, 30)| ├── 3.npy # the npy file represents the visual knowledge corresponding to the region with ID 3， shape (num_SVI_3, 30)| ├── imgs/ | ├── 1.npy # the npy file represents the street view imagery corresponding to the region with ID 1， shape (num_SVI_1, 3, 256, 256)| ├── 2.npy # the npy file represents the street view imagery corresponding to the region with ID 2， shape (num_SVI_2, 3, 256, 256)| ├── 3.npy # the npy file represents the street view imagery corresponding to the region with ID 3， shape (num_SVI_3, 3, 256, 256)| ├── fliter/ | ├── train_flow.npy # the taxi flow count of the training set, shape: (n_region_train,5)| ├── train_id.npy # the regions of the training set, shape: (n_region_train,)| ├── val_flow.npy # the taxi flow count of the validation set, shape: (n_region_val,5)| ├── val_id.npy # the regions of the validation set, shape: (n_region_val,)with: n: the number of image samples n_train: the number of training samples n_val: the number of validation samples n_train : n_val = 7 : 3 n_region_train: the number of regions in the training set n_region_val: the number of regions in the validation set n_region_train : n_region_val = 7 : 3 num_SVI_1: the number of street view images collected in the region with id 1 num_SVI_2: the number of street view images collected in the region with id 2 num_SVI_3: the number of street view images collected in the region with id 3For detailed descriptions of the data, its format, and how to use it in the experiment, please refer to the README-data.md file.

Collection of static general Bloomington and city street maps of various sizes.

The Digital City Map (DCM) data represents street lines and other features shown on the City Map, which is the official street map of the City of New York. The City Map consists of 5 different sets of maps, one for each borough, totaling over 8000 individual paper maps. The DCM datasets were created in an ongoing effort to digitize official street records and bring them together with other street information to make them easily accessible to the public. The Digital City Map (DCM) is comprised of seven datasets; Digital City Map, Street Center Line, City Map Alterations, Arterial Highways and Major Streets, Street Name Changes (areas), Street Name Changes (lines), and Street Name Changes (points). All of the Digital City Map (DCM) datasets are featured on the Streets App All previously released versions of this data are available at BYTES of the BIG APPLE- Archive