OpenCV Python Tutorial #7 - Template Matching (Object Detection)

Welcome to another OpenCV tutorial! We'll be talking about template matching (object detection). In this video, we'll use an image of a soccer practice and use OpenCV to detect the soccer ball and find it's exact location. For example, the image of the ball will be our template, and we'll use OpenCV to match it.

⭐️ Timestamps ⭐️
00:00 | Setup & Images
02:35 | Loading Template & Base Images
04:55 | Template Matching Methods
07:00 | Theory Behind Template Matching
14:10 | Displaying Matches

◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️
💰 Courses & Merch 💰

🔗 Social Medias 🔗

🎬 My YouTube Gear 🎬

💸 Donations 💸
◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️

⭐️ Tags ⭐️
- Object detection in OpenCV
- Template matching OpenCV
- Python
- Computer vision
- Cameras and video capture

⭐️ Hashtags ⭐️
#OpenCV #Python #TemplateMatching

[ananthramvijayaraj]

Tim you don't need to apologize for the lengthy explanation, that's the whole reason why we are here.

[Atharva_Vashist]

I was interested in coding and chose python as my language to start, and you helped me to clear my basics, and explore many modules like opencv, kivy etc. You explain very well. Thank you for your videos. You are the best youtuber teaching python. Keep it

[JordanFayter]

Tim you do an amazing job no need to apologize for the lengthy explanations they are actually extremely useful. I'll usually watch your videos multiple times just so I know I'm getting the information.

[raviolismudge]

FOR THE LOVE OF GOD, I SPENT MY WHOLE DAY ON THIS T-T
Thank you so much. I am crying tears of joy.
My professor refuse to help us with anything that have to do with programming, he only talk about the math behind visualization.
Thank you so much.

[DL-yujg]

Gotta say, this is one of the most well explained Python-related lectures I've ever witnessed. Thank you so much for the video. Really helped me A LOT.

[Traxirage]

Great tutorial!

For future reference, if you want to select a segment of two lines, you can hold Shift + Alt and drag the mouse. This way it won't select all the way to the end of row, but only to the column you move your cursor to.

[jaakkaan]

AMAZING VIDEO THANK YOU for the thoroughness. Absolutely covered all areas where confusion or curiosity could come up!

[sunwukong]

Your "lengthy" explanation was great. Not as long as you might think and comprehensable.

[alexandrevalente]

Great example to illustrate the use of the convolution here.

[Ferret]

Thanks for this series Tim. You have a very nice style of presenting and explaining how the code works.

[endab]

You are gifted. Thank you for the content.

[mehmetsezgin]

Tim that is pretty straightforward tutorial. Thanks !

[nicolasa.bermellferrer]

Your explanation was actually pretty good. Thank you!!

[J-n]

Your tutors are the best, when you don't show your distracting face or when I cover it with black rectangle. I guess they are best for learning Python syntax.

[phamminhhung]

Actually your long explanation is useful for us, it illuminates us how things work and you should keep doing that, or maybe you can implement some drawings

[tsraikage]

how sad I've never found your channel before. i would be step or two higher than i am right now. your explanation methods are outstanding. great job

[hackercop]

Glitch at 12:20 that he forgot to edit out lol. Great video, am trying to learn object recognition.

[RandomGuy-dfoy]

At the end of the video, you don't need to resize the image at every step. Just resize where you are showing the image at line 23 as
cv2.imshow("Match", cv2.resize(img2, (0, 0), fx=0.8, fy=0.8))

[TeckWill]

Great job with your tutorial videos Tim! I understand that with the methods you shown we can find the exact same pixels pattern in the image. Now, I wonder, for example, how could we find all the shoes on the image from the template image without them being perfectly identical. I suppose it will be necessary to go towards artificial intelligence / deep learning? I'd like to know what you would do and see a tutorial video on it.

[aimanyounis]

template mathcing is not convolution it's correlation, which means that you don't flip the kernel before you pass it on the whole image : another defintion of correlation is convolution withour flipping.