Annotation of Live Video Streams for Traffic Management and Road Planning

Annotation Apr 20th, 2024

The Objective:

The client a data analytical company providing solutions to government agencies was looking to categorize and label huge number of vehicles based on movement like approach, turning movement etc for the following reasons:

  • To assist department of Civil and Environmental Engineering in developing proper road plans for smooth traffic movement.
  • To develop machine learning solution predicting traffic related issues like congestion, accident prevention, better road planning, lane movement etc.
  • To train client’s machine learning algorithms through labelled videos with added metadata to make the objects identifiable in videos.
  • To develop models and technology to evaluate if video analytics can track traffic situation from live video feeds.

Therefore, the client partnered with HabileData to label vehicle images in pre-defined criteria.

The Challenges:

  • Hiring skilled workforce with experience in building complex computer vision models related to standard automobile classification.
  • Skilled annotators who could annotate images from videos under different lighting, weather and erratic traffic volumes.
  • Dividing the workforce into shifts to train massive amount of data.
  • Expertise in labelling images from traffic videos as often the videos became unsteady with blurred images, obstructions etc.
  • Choosing the right annotation method and accurately labelling the data required to train the machine how to recognize them visually, like a person.
  • Managing the complex process of annotating videos or multi-frame data.
  • Counting pedestrians and bicycles from in-pavement loops not differentiated for directions of movement (going straight and turning right).

HabileData’s Solution

  • Workflow was planned to categorize and label huge volumes of vehicle and pedestrian images from live as well as video feeds from across major cities in US and Canada.
  • The annotated images were used as training data for machine learning models.
  • A well-documented 5-step process helped successfully annotate image labelling
  • Data was sourced in two forms:
    • Pre-recorded videos
    • Live video streams
  • Credentials to log into the City’s traffic camera network gave annotators access to live video streams.
  • Labelling and segmentation was done as per the following norms:
    • Vehicles labelled by – category, model name, colour and direction of vehicle
    • 14 categories included – Car, SUV, small truck, medium truck, large truck, pedestrian, bus, van, group of people, bicycle, motorcycle, traffic signal-green, traffic signal yellow, and traffic signal-red.
    • Vehicles were classified tagged and segmented by turning movement or by the direction of approach
    • Obstructed vehicles were not labelled
    • Any ambiguity on the vehicle due to poor light or weather conditions were re-validated by the client
  • Count of vehicles in individual lanes was done through line-based technique.
  • State of the line changed from unoccupied to occupied and then back to unoccupied increasing the count of said line.
  • A demarcation line (red line) was used to demarcate small vehicle that was not labelled.
  • A team of senior auditors audited around 10% of the annotated images.
  • Any anomalies or deviations in the data were used for training purpose.
  • Any annotation found erroneous was taken up for re-labelling.
  • City wise segregation helped in uploading the labelled images on OneDrive.
  • Report listing number and types of vehicles annotated was generated for record purpose.

Business Impact

  • Huge volumes of training data to power machine learning
  • A dashboard of directional traffic volumes providing live data and alerts
  • Video analytical solution for traffic developed

Value Addition

Annotating pre-recorded and live video stream of vehicles provided training data for machine learning models for a California based data analytics company helped managing traffic efficiently.

Go to Top