Skip to content

This dataset could be used for automatically finding and categorizing potholes in city streets so the worst ones can be fixed faster.

Notifications You must be signed in to change notification settings

Tanwar-12/POT-HOLE-DETECTION-YOLOV5

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

pothole_video.mp4

POT-HOLE-DETECTION-YOLOV5

Overview :

The purpose of this project is to create a Deep Learning model and this dataset could be used for automatically finding and categorizing potholes in city streets so the worst ones can be fixed faster.

📁 Dataset Used :

https://public.roboflow.com/object-detection/pothole

This dataset consist of only one class: (Pothole)

Workflow:

Data Preparation:

  • Total 465 images for training and 133 images for validation present in 2 classes.
  • Create a bounding boxes with the help of label-img And makesense.ai website according to YoloV5.
  • Prepare folder structure that can be accept by YoloV5. train folders

Steps to use Yolov5:

  • Cloning the YoloV5 file from official repository.
  • Changing the directory of yolov5
  • Installing the dependencies
  • Download all versions pre-trained weights.

Steps Before Training Custom Dataset:

  • Go to yolov5/data/.
  • Open data.yaml
  • Edit the following inside it:
  1. Training and Validation file path
  2. Number of classes and Class names.

Training YOLOV5 Model

  • Set images size 640 with batch of 8.
  • Train model around 600 epochs .
  • Visualise the training metrics with the help of tensorboard.

Testing Images Using Test Data

image

Testing Video Demo:

Demo.Pothole.mp4

Just follow☝️ me and Star⭐ my repository

Motivated and supported by the works of,

[Miss. Sakshi Tanwar].

About

This dataset could be used for automatically finding and categorizing potholes in city streets so the worst ones can be fixed faster.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published