0

I am new to the Object Detection world. So don't mind my basic question.

I have an Identycard Image. That Image contains the portable IDCard and other information related to the Card Holder.

Currently, I just want to crop out the IDCard.

Before, I used to crop this programmatically as I never thought that the Position of IDCard could be changed.

The Image Looks like this:

enter image description here I want to crop out the IDCard area at the bottom.

As I stated earlier that the position, Width, and Height of IDCard are not fixed.

Now my main question is that Should I go for ML to make this work, Would it be overkill?

If it's not then what are the steps that I can follow to detect the IDCard at bottom starting from the cut from here part till bottom. I don't have any ML Knowledge right now.

If it's indeed a overkill than what could be the way to detect the card?

ChandraShekharAazad
  • 91
  • 7

1 Answers1

0

ML is an overkill here. If the position changes but the pixels that mark the line to cut stay the same, a search for these pixels is the way to go.

pixels

r3dapple
  • 431
  • 5
  • 16
  • Can you explain about the overkill thing. – ChandraShekharAazad Feb 05 '21 at 13:39
  • Its just not necessary to use machine learning here. The pixel search algorithm is a lot easier, probably faster and for sure more reliable. – r3dapple Feb 05 '21 at 13:48
  • This is not only the image that i want to crop id card from there are others too which may not include the scissors part. Using ML to detect id type and boundingBoxes was my long term plan. Is there any information you could share that can make this thing a bit easier to follow ?. – ChandraShekharAazad Feb 06 '21 at 14:50
  • Sure. Can you share a couple more pictures and mark on them where they are supposed to be cut? – r3dapple Feb 06 '21 at 20:40
  • can you please continue on this topic ? i want to know more about pixel search and Custom object detection both. This will be really helpfull. still i am stuck at finding the WAY to do it. – ChandraShekharAazad Mar 18 '21 at 12:19
  • This might be helpful to you:https://stackoverflow.com/a/18075751/9298528 As for the algorithm itself, you're gonna have to come with that yourself. Its really simple. – r3dapple Mar 18 '21 at 19:00
  • hello mate, can you look at this one ? https://stackoverflow.com/questions/67498958/opencv-js-template-matching-speed-and-using-multiple-images – ChandraShekharAazad May 15 '21 at 05:51