INFRARED Computer Vision Project

Drop an image or

This infrared street scenes computer vision dataset labels cars, people, bicycles, and dogs (although dogs are severely under-represented; see the health check for more info on class distribution). It contains over 10,000 grayscaled thermal images taken with a FLIR camera that can be used for detecting objects at night or in poor lighting conditions.

It may be particularly useful in creating self-driving autonomous vehicles, due to the forward-looking dashcam style footage of roads, pedestrians, and vehicles. There is also a pre-trained model available for testing.

Example use-cases include:

  • Prototyping a self-driving vehicle
  • Creating a tool that warns drivers if pedestrians are about to cross their path
  • Counting cars driving by (or people walking by) at night
  • Intelligently turning lights on and off for people and vehicles (but not animals)
  • Detecting unauthorized access to unlit facilities

You could also re-label other objects of interest from this dataset to train a model to detect other types of things.

Cite this Project

If you use this dataset in a research paper, please cite it using the following BibTeX:

@misc{ infrared-6s0ke_dataset,
    title = { INFRARED Dataset },
    type = { Open Source Dataset },
    author = { Calib Buckton },
    howpublished = { \url{ } },
    url = { },
    journal = { Roboflow Universe },
    publisher = { Roboflow },
    year = { 2022 },
    month = { sep },
    note = { visited on 2022-10-02 },


Calib Buckton

Last Updated

13 days ago

Project Type

Object Detection




bicycle, car, dog, person


CC BY 4.0