metadata

task_categories:
  - object-detection
language:
  - en
tags:
  - computer vision
  - code
  - python
  - traffic
  - singapore
  - roadway
pretty_name: Traffic Images for Object Detection
size_categories:
  - 10K<n<100K

Traffic Image Data Extraction Through Singapore Government API

Description

The Singapore government offers real-time images from traffic cameras across the nation through its API. This dataset compiles a comprehensive image dataset in the form of a DataFrame by extracting data for the month of January 2024 from 6 pm to 7 pm each day using the API.

Below are sample images from the dataset:

Use Cases

The resulting dataset will facilitate easy integration into various use cases including:

Object Detection

Utilize the dataset for training object detection models to identify and analyze vehicles, pedestrians, and other objects in the traffic images.

Traffic Trend Analysis

Leverage time-series analysis to identify and analyze traffic trends over specific periods. This can provide valuable insights into peak traffic times, congestion patterns, and potential areas for infrastructure improvement.

Road Safety Assessment

Implement computer vision algorithms to assess road safety by analyzing traffic images for potential hazards, unusual road conditions, or non-compliance with traffic rules. This use case aims to enhance road safety monitoring and contribute to the development of intelligent transportation systems.

Dataset Details

The dataset will comprise the following columns:

Timestamp: Date and time of the image acquisition from LTA's Datamall.
Camera_ID: Unique identifier assigned by LTA to each traffic camera.
Latitude: Geographic coordinate of the camera's location (latitude).
Longitude: Geographic coordinate of the camera's location (longitude).
Image_URL: The traffic image fetched from the Image_URL provided by the API.
Image_Metadata: Metadata of the image file including height, width, and MD5 hash.

Limitations of my Dataset

The Dataset due to limited computational capability has data of only one month and 1 hour for each day. Fetching large data (such as a year) would help in analysing the macro trends and significant patterns.

API Documentation

For more details on accessing the traffic camera images, visit the API Documentation.

Use Case

Refer to the attached traffic_object_detection.py file to see how I used a pretrained YOLO model to detech cars and trucks. Further I generated traffic insights using an interactive streamlit dashboard (code not on HuggingFace).

Below is a sample output of the YOLO model Sample image from the data

Here are the snippets of my Dashboard:

Version 2.0 of the dataset and analysis coming soon!