- Main Idea Behind the Project
I had some personal interest on computer vision as they are much helpful to do image processing and these kinds of stuffs lot easily.So I decided to make something that may fulfill my wish to learn about CV2(computer vision version 2.0)and also a bit of image processing.So I was hoping to get a perfect time slot and some free amount of time to learn all these.Finally I got time on last year's long 2 month vacation and fully utilized it for my learning purpose.Actually for that reason I can tell now I can write and share this project to you.
- What it does ?
The module can detect only two types of shapes : rectangle and square from the live video.So we used the webcam of the laptop and used that to capture live videos and identify any rectangular shaped object using open CV.
- Platform we used
I used basic Python Shell scripts for scripting and some open CV2 libraries were required to run them.I could have used "Anaconda" or "Pycharm" also but I felt rather easy manually doing those stuffs by feeling the necessity of every module that I included while needed.
- Difficulties we faced
Firstly I would like to mention that learning Python and Open CV was really really fun but to be honest, making the accuracy up to 100% is really really challenging. So I tried to maximize those as much as I can but there are other issues like brightness, contrast between shadow and light.So for these I have to go through several filters.I had to make them firstly Gaussian Blur so that not too much intense light can affect the performance but still there are lack of accuracy but it can predict most of the time.I also faced problem to analyze the minimum threshold(in my case,the the minimal area that we can set as a rectangle).Finally after some trial and error test I got it suitable for my case as 500 ~ 100000.
- Some moments
The whole project code can be found in Github.Please follow the link here.
No comments:
Post a Comment