How AI Learns to See Images | Computer Vision Explained #ai #computervision

本站添加: 2026-05-25

[00:00:00]In the previous reel, we discussed [music] how computer vision does not work with simple if-else rules. So, how does computer vision learn from images?

[00:00:06]Not by rules, not by magic, by making mistakes, millions of them. Imagine training an AI to tell [music] apple from oranges. At first, it knows nothing. You show it an apple, it guesses apple 40% orange 60%. So, the answer is wrong. We tell [music] it that was an apple. But, here's the important part. AI does not understand the mistakes like [music] humans do. It calculates how far its answer was from the correct answer. This difference is called loss. The bigger the mistake, the higher the loss. Then, it slightly adjusts its internal settings, sees the next image, >> [music] >> predict again, check the mistake. This process repeats thousands of times.

[00:00:39]Slowly, it starts noticing patterns, roundness, color, texture, stem shape.

[00:00:43]After enough example, it sees a new fruit [music] and says 92% apple. And that's it. That's the whole trick. That is how AI learns, [music] not by if-else, not by magic, but by repeated corrections. Like and subscribe Ishan for you if you want computer vision explained in simple language.