Pothole Detection

Potholes are the lowly annoyance that in fact cause billions of dollars in vehicle damages and a large portion of highway deaths. Of approximately 33,000 traffic fatalities each year, one-third involve poor road conditions. The grave danger that potholes present to bicyclists can also be costly to municipalities that fail to fix hazardous potholes and face multi-million dollar lawsuits as a result.

 

TECHNOLOGY
Python
Numpy & Pandas
Tensorflow
OpenCV

Key Challenges

  • As the cost of procuring video feed from multiple devices is large, so solution should filter out probable pothole locations.
  • High noise in sensor data adds additional challenge of cleanup and filtering.


Our Solutions

  • Pattern and Anomaly Analysis helped in creating more valuable and actionable insights from sensor data.
  • Transfer Learning techniques applied on Deep Learning Neural Networks helps detect potholes in data.
  • Advanced Vision segmentation is applied to mark the potholes on the road along with precise location.