Measuring Fog Dispersal with JPEG File Sizes
The author uses a zero-cost sensor system based on JPEG file sizes from a webcam to track changes in the environment, specifically the dispersal of fog over time. The article explores the insights and limitations of this approach.
Why it matters
This novel approach to environmental sensing using JPEG file sizes provides insights into the dynamics of fog dispersal, demonstrating the potential for low-cost, creative solutions in AI and machine learning.
Key Points
- 1The author's sensor system uses JPEG file sizes to infer scene information density, with smaller files indicating dense fog and larger files indicating clear skies
- 2The sensor captured a 104% increase in file size as the fog cleared, but the exact rate of dispersal is unknown due to the 5-hour gap between observations
- 3Comparison of data from different days shows that while the morning fog conditions vary, the afternoon clear sky converges to a consistent file size
- 4The author plans to run a higher-resolution timelapse to better understand the fog dispersal process
Details
The author has been running a zero-cost environmental perception system on a 2014 MacBook, using the Photo Booth app and JPEG file sizes to track changes in the environment. The insight is that JPEG compression output correlates with scene information density, with fog eliminating visual contrast and detail resulting in smaller files, and clear skies with buildings and trees resulting in larger files. \n\nOn April 18th, the sensor recorded a 104% increase in file size as the fog cleared, but the exact rate of dispersal is unknown due to the 5-hour gap between observations. The author compares this to data from April 16th, which showed a more gradual dispersal pattern over 53 minutes. \n\nThe article explores the deeper pattern observed across 9 days of data, noting that while the morning fog conditions vary, the afternoon clear sky converges to a consistent file size of around 205 KB. The author concludes that
No comments yet
Be the first to comment