Taxi Availability Singapore

We're interested in seeing how traffic levels have changed between Jan 2019 - a normal year, and Jan 2020 - a COVID year in Singapore.

The hypothesis is that the levels have dropped, and we're determined to find out through looking at the traffic cameras and computer vision.

📝 Table of Contents

🗂 Data

The data is from Singapore government provided API at https://data.gov.sg/dataset/traffic-images .

Flow of project

  1. Explore Singapore's traffic camera API and map the locations of the cameras (Part I)
  2. Make two one-month API calls to collect the initial data (Part II)
  3. Collecting massive amounts of image data from API (Part III)
  4. Prepare for GPU-enabled object detection (Part III)
  5. Using computer vision to count traffic in 2019 and 2020 (Part IV)
  6. Compare the traffic between two years (Part V)

Part 1

  • Explore the traffic camera API from Singapore's Data.gov.sg
  • Map the locations of the traffic camera in Singapore
  • Identify the best camera to focus on for data collection

Part 2

  • Collect two years' worth of traffic image JSON
  • Combine the eventual DataFrame
  • Filter for camera ID 1709 only
  • Export DataFrame for Part III

Part 3

  • Write a function to download images into drive
  • Execute the function concurrently
  • Prepare for OpenCV GPU execution in Part IV

Part 4

  • Test object detection on one image
  • Detect number of objects in all images
  • Updating the DataFrames with the vehicle count

Part 5

  • Transform and simplify the DataFrames
  • Plot the respective num_cars over time

🎉 Acknowledgments

  • Tribe Academy