Datasets

    Monocular road traffic images dataset

  • Download

  • Readme: each folder will contain a trip with pairs of images and text files. Their correspondence is set by having the same name (UNIX timestamp). The images are 640x480 px and the text files contain the following data:
    • timestamp (UNIX timestamp)
    • accX, accY, accZ (accelerometer sensor data on 3 axis) - m/s^2
    • gyroX, gyroY, gyroZ (gyroscope sensor data on 3 axis) - rad/s
    • magX, magY, magZ (geomagnetic field sensor data on 3 axis) - rad
    • yaw, pitch, roll angles - rad
    • yawRate (same ass gyroX) - rad/s
    • speedMs (speed) - m/s
    • lat, lng (latitude and longitude coordinates)

    The dataset also contains the segmented images having the "_mask" suffix.

    Cite paper:
  • Razvan Itu, Radu Danescu, "A Self-Calibrating Probabilistic Framework for 3D Environment Perception Using Monocular Vision", Sensors, Vol. 5, No. 20, 2020, Art. No. 1280.

  • Monocular Perception system video demo


    Vanishing Point from traffic images dataset

  • Download (readme file is included in the zip archive)

  • Cite paper:
  • Razvan Itu, Diana Borza, Radu Danescu, "Automatic extrinsic camera parameters calibration using Convolutional Neural Networks", 2017 IEEE 13th International Conference on Intelligent Computer Communication and Processing (ICCP 2017), pp. 273-278.

  • IEEE Explore: here