DeepOSM

Visit Tool

DeepOSM is an Open Source Data & Analytics tool that classifies roads and features in satellite imagery by training neural networks with OpenStreetMap (OSM) data. It can download satellite imagery and OSM data, generate training data, and display predictions.

Claim this tool

No Views Yet

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is DeepOSM?

DeepOSM is an open-source tool designed for training deep learning networks to classify roads and features within satellite imagery, leveraging OpenStreetMap (OSM) data. It streamlines the process by allowing users to download chunks of satellite imagery and corresponding OSM data for specific areas. The tool then generates training and evaluation data, facilitating the development and testing of neural networks. DeepOSM can display predictions of mis-registered roads in OSM data or provide raw ON/OFF predictions, making it valuable for improving the accuracy and completeness of mapping data. It is designed to be run with Docker and supports GPU acceleration for faster processing.

Best used for

Ideal for data scientists and researchers who need to train deep learning models for geospatial analysis, identify mis-registered roads in OpenStreetMap data, and generate training data from satellite imagery. Especially valuable for enhancing the accuracy and completeness of mapping information.

Common actions

analyze geospatial data

train deep learning models

classify satellite imagery

improve mapping data

low-code/no-codeworkflowsopen-sourceautomated workflowcollaborationdeepfake"AI Agents"github copilotface swapping

Capabilities

Key features

Classify roads/features
Download satellite imagery
Download OSM data
Generate training data
Display mis-registered road predictions
GPU acceleration support
Docker integration

Target Audience

data scientist

Integrations

Not yet documented

Pricing & Plans

Open Source

Free

FAQs

What kind of satellite imagery does DeepOSM use for training?

DeepOSM primarily uses NAIP (National Agriculture Imagery Program) images, which provide 1-meter-per-pixel resolution with RGB+infrared data bands. These images are sourced from a requester-pays S3 bucket set up by Mapbox, requiring AWS credentials for access.

What are the system requirements to run DeepOSM?

DeepOSM has been successfully run on Mac (10.x) and Linux (14.04 and 16.04). It requires at least 4GB of memory. For GPU acceleration, NVIDIA drivers and nvidia-docker must be installed, along with a compatible GPU.

How accurate are DeepOSM's predictions by default?

By default, DeepOSM can achieve 75-80% overall accuracy when predicting if the center 9px of a 64px tile contains a road, using infrared and RGB bands. This is achieved with a single fully-connected ReLU layer in TensorFlow, training for only about a minute.

Trending

Subcategories trending in Data & Analytics

Business Intelligence Predictive Analytics Data Labeling & Annotation Real-Time Analytics Market Research Data Cleaning & Prep

Trending

Explore

Browse AI tools by category

Content & Design Productivity & Business Coding & Development AI Agents & Automation Research & Education Wellness & Lifestyle Career Development Marketing & Growth Data & Analytics Customer Support & CX Finance E-commerce