/Wonder_Vision_Lane_Detection

This repository is dedicated for showing the work of the "Lane Detection" subgroup within #sg_wonder_vision channel on Slack Work-space for Private and Secure AI challenge from Udacity

Primary LanguageJupyter Notebook

Wonder_Vision_Lane_Detection

This repository is dedicated for showing the work of the "Lane Detection" subgroup within #sg_wonder_vision channel on Slack Work-space for Private and Secure AI challenge from Udacity.

Project overview

As the name suggests, this project focusses on identification of lanes of the road which we achieved through this project using various approaches and frameworks including but not limited to OpenCV, PyTorch etc. This model is to be used in drones , self driving cars, surveillance systems , so its is primarily related to 'Safety'.

Approaches followed

  1. Reading Images
  2. Color Filtering in HLS
  3. Region of Interest
  4. Canny Edge Detection
  5. Hough Line Detection
  6. Line Filtering & Averaging
  7. Overlay detected lane
  8. Applying to Video

Outcome of the project

2019-08-20 16_41_50-Slack _ Mohammad Diab, Shudipto Trafder, shivu, Astha Adhikari, Droid, Oudarjya

2019-08-20 16_41_44-Slack _ Mohammad Diab, Shudipto Trafder, shivu, Astha Adhikari, Droid, Oudarjya

Project Participants

Following are the Slack handles of the members

Name Slack Name Github
Astha Adhikari @Astha Adhikari https://github.com/adhikariastha5
Oudarjya Sen Sarma @Oudarjya Sen Sarma https://github.com/oudarjya718
Mohammad Diab @Mohammad Diab https://github.com/depo-egy
Shiva Shankar @shivu https://github.com/shiv-u
Vigneshwari Ramakrishnan @Vigneshwari https://github.com/drvigneshwari

This Google document is the for meetups follow up:

https://docs.google.com/document/d/1fGN4T_ZJNQP5KHoIm5kUwVwMsGtgJyX9qYuI5ZCVQWc/edit

Here's the Workflow of our project!!!

BlockDiagramLast

Use case of our Lane Dection Project:

  • It can be used in drones We can use lane detection in automatic drones which are used nowadays like medical,traffic control,monitoring and so on.
  • It can be used in air surveillance system. We can use this lane detection in to get a survey of places affected by disasters or in general normal surveillance.
  • We can even measure the traffic crowd in particlular lane.

Using Pytorch Transfer Learning for car detection:

We tried using vgg model for detecting the car using transfer learning. Though we could not completely integrate it with our project, we have a kept a small notebook for the things we have tried.

Yolo object detection in video:

We also tried the object detection in a video using yolo and the weights are in the drive link above.

Our Future Enhancement(Our Ultimate Goal):

  • Detecting urban and rural roads.
  • Use car detection in roads for a particular lane and count frequency (real time video)
  • Use this detection with other projects in #sg_wonder_vision channel to create a mass useful project.