OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Sam Road

Sam Road is an open-source Python computer-vision codebase built around Segment Anything Model for extracting road-network graphs and masks from aerial imagery. It includes workflows for training on CityScale or SpaceNet, running inference, and evaluating results with APLS and TOPO metrics.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Sam Road — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Sam Road when a research or mapping workflow needs road-network graphs from aerial imagery and can provide the required GPU, checkpoint and dataset setup.

Who it is for

  • Remote-sensing and computer-vision researchers working on road extraction
  • Mapping teams exploring graph-based road-network extraction from aerial imagery
  • Developers who can run Python model code with a CUDA-capable Nvidia GPU

Documented uses

  • Train a road-extraction model on CityScale or SpaceNet data.
  • Produce road-network graph predictions, masks and visualizations from aerial imagery.
  • Evaluate road-network outputs using APLS and TOPO.
Before you choose

Requirements & limits

The repository is public and lists an MIT license. The code is run in a local or server environment; the documented requirements include an Nvidia GPU with CUDA.

  • The documented installation requires an Nvidia GPU with CUDA and its driver.
  • Training and inference depend on a ViT-B SAM checkpoint and prepared CityScale or SpaceNet data.
  • The README specifies Go for APLS evaluation; this requirement is separate from the stated model inference command.
  • The workflow uses PyTorch, PyTorch Lightning and Weights & Biases.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run inference with a prepared checkpoint

Generate road-network inference results and visualizations from a prepared dataset workflow.

Before you start

  • The Sam Road codebase and its documented PyTorch dependencies
  • An Nvidia GPU with CUDA and driver
  • A prepared dataset and a trained or otherwise available checkpoint
  • The configuration corresponding to the checkpoint

Steps

  1. Select the configuration used for training, such as config/toponet_vitb_512_cityscale.yaml or config/toponet_vitb_256_spacenet.yaml.

  2. Run python inferencer.py --config=path_to_the_same_config_for_training --checkpoint=path_to_ckpt, replacing the placeholders with the matching configuration and checkpoint paths.

Expected outcome

Inference results and visualizations are saved by the script.

Guide limitations

  • The README requires the configuration to match the training workflow.
  • The command assumes the checkpoint and dataset setup are already available.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
htcr/sam_road · Official homepage title · retained in its source language
How we review sourcesSuggest a correction