๋ณธ๋ฌธ ๋ฐ”๋กœ๊ฐ€๊ธฐ
  • ONI์™€ Jean-y์˜ ์Šค๋งˆํŠธํ•œ ๊ณต๊ฐ„ ๐Ÿ”Ž

๐Ÿค– AI์ผ์ž˜๋Ÿฌ ์˜ค๋‹ˆ์˜ AI11

Canny edge detection์ด๋ž€? ์•ˆ๋…•ํ•˜์„ธ์š” AI์ผ์ž˜๋Ÿฌ ์˜ค๋‹ˆ์ž…๋‹ˆ๋‹ค ๐Ÿค– ์˜ค๋Š˜์€ Canny edge detection์— ๋Œ€ํ•œ ๊ธ€์„ ์ž‘์„ฑํ•ด๋ณด์•˜์Šต๋‹ˆ๋‹ค ๐Ÿ–‹ ๊ทธ๋Ÿผ Canny edge detection์— ๋Œ€ํ•ด ์•Œ์•„๋ณด๋Š” ์‹œ๊ฐ„์„ ๊ฐ€์ ธ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค ๐Ÿ‘Œ๐Ÿป The Canny edge detector is a popular algorithm for detecting edges in an image. It is known for its ability to accurately detect edges while minimizing false detections. - Computer Vision: Algorithms and Applications by Richard Szeliski โœ” ๋“ค์–ด๊ฐ€๋ฉฐ ์ปดํ“จํ„ฐ ๋น„์ „์— ๊ด€์‹ฌ์ด ์žˆ๊ฑฐ๋‚˜ ๋จธ์‹  ๋Ÿฌ๋‹ ์‹ค๋ฌด์ž๋ผ๋ฉด Canny edge dete.. 2023. 3. 28.
๋จธ์‹ ๋น„์ „(Machine Vision)์ด๋ž€? ์•ˆ๋…•ํ•˜์„ธ์š” AI์ผ์ž˜๋Ÿฌ ์˜ค๋‹ˆ์ž…๋‹ˆ๋‹ค ๐Ÿค– ์˜ค๋Š˜์€ ๋จธ์‹  ๋น„์ „์— ๋Œ€ํ•œ ๊ธ€์„ ์ž‘์„ฑํ•ด๋ณด์•˜์Šต๋‹ˆ๋‹ค ๐Ÿ–‹ ๊ทธ๋Ÿผ ๋จธ์‹  ๋น„์ „์— ๋Œ€ํ•ด ์•Œ์•„๋ณด๋Š” ์‹œ๊ฐ„์„ ๊ฐ€์ ธ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค ๐Ÿ‘Œ๐Ÿป Machine vision is a key technology for Industry 4.0, enabling machines to see and interpret the world around them with unprecedented accuracy and speed. - Dr. Christian Vollrath, CEO of Vision Components โœ” ๋“ค์–ด๊ฐ€๋ฉฐ ๊ธฐ์ˆ ์ด ๋ฐœ์ „ํ•จ์— ๋”ฐ๋ผ ์ด๋ฏธ์ง€๋ฅผ ์บก์ฒ˜ํ•˜๊ณ  ๋ถ„์„ํ•˜๋Š” ๊ธฐ๋Šฅ๋„ ๋ฐœ์ „ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ปดํ“จํ„ฐ ๋น„์ „์ด๋ผ๊ณ ๋„ ํ•˜๋Š” ๋จธ์‹  ๋น„์ „์€ ๊ธฐ๊ณ„์— ์‹œ๊ฐ ์ •๋ณด๋ฅผ ํ•ด์„ํ•˜๊ณ  ์ดํ•ดํ•˜๋Š” ๋Šฅ๋ ฅ์„ ์ œ๊ณตํ•˜๋Š” ๋ฐ ์ค‘์ ์„ ๋‘” ์ธ๊ณต ์ง€๋Šฅ.. 2023. 3. 25.
Gradient Descent(๊ฒฝ์‚ฌํ•˜๊ฐ•๋ฒ•)์ด๋ž€? ์•ˆ๋…•ํ•˜์„ธ์š” AI์ผ์ž˜๋Ÿฌ ์˜ค๋‹ˆ์ž…๋‹ˆ๋‹ค ๐Ÿค– ์˜ค๋Š˜์€ Gradient Descent์— ๋Œ€ํ•œ ๊ธ€์„ ์ž‘์„ฑํ•ด๋ณด์•˜์Šต๋‹ˆ๋‹ค ๐Ÿ–‹ ๊ทธ๋Ÿผ Gradient Descent์— ๋Œ€ํ•ด ์•Œ์•„๋ณด๋Š” ์‹œ๊ฐ„์„ ๊ฐ€์ ธ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค ๐Ÿ‘Œ๐Ÿป Gradient descent is like a ball rolling down a hill. The steeper the hill, the faster the ball will roll. Likewise, the steeper the gradient, the faster the convergence. - Andrew Ng โœ” ๋“ค์–ด๊ฐ€๋ฉฐ ๋จธ์‹  ๋Ÿฌ๋‹์— ๊ด€์‹ฌ์ด ์žˆ๋‹ค๋ฉด "Gradient descent"์ด๋ผ๋Š” ์šฉ์–ด๋ฅผ ๋“ค์–ด๋ณด์…จ์„ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๊ฒฝ์‚ฌ ํ•˜๊ฐ•์€ ๋จธ์‹  ๋Ÿฌ๋‹๊ณผ ๋”ฅ ๋Ÿฌ๋‹์—์„œ ๋น„์šฉ ํ•จ์ˆ˜ ๋˜๋Š” ์†์‹ค ํ•จ์ˆ˜๋ฅผ ์ตœ์†Œํ™”ํ•˜๊ธฐ ์œ„ํ•ด ๋„๋ฆฌ ์‚ฌ์šฉ๋˜๋Š”.. 2023. 3. 24.
์ด๋ฏธ์ง€๋„ท(ImageNet) ์ด๋ž€? ์•ˆ๋…•ํ•˜์„ธ์š” AI์ผ์ž˜๋Ÿฌ ์˜ค๋‹ˆ์ž…๋‹ˆ๋‹ค ๐Ÿค– ์˜ค๋Š˜์€ ์ด๋ฏธ์ง€๋„ท์— ๋Œ€ํ•œ ๊ธ€์„ ์ž‘์„ฑํ•ด๋ณด์•˜์Šต๋‹ˆ๋‹ค ๐Ÿ–‹ ๊ทธ๋Ÿผ ์ด๋ฏธ์ง€๋„ท์— ๋Œ€ํ•ด ์•Œ์•„๋ณด๋Š” ์‹œ๊ฐ„์„ ๊ฐ€์ ธ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค ๐Ÿ‘Œ๐Ÿป The power of ImageNet lies not just in the vast amount of data it contains, but in the insights it provides into the workings of deep learning models. - Yann LeCun, Director of AI Research at Facebook โœ” ๋“ค์–ด๊ฐ€๋ฉฐ ์ตœ๊ทผ ๋ช‡ ๋…„ ๋™์•ˆ ์ปดํ“จํ„ฐ ๋น„์ „์€ ๋”ฅ๋Ÿฌ๋‹ ์•Œ๊ณ ๋ฆฌ์ฆ˜์˜ ๊ฐœ๋ฐœ๊ณผ ๋Œ€๊ทœ๋ชจ ๋ฐ์ดํ„ฐ ์„ธํŠธ์˜ ๊ฐ€์šฉ์„ฑ์œผ๋กœ ์ธํ•ด ๊ธ‰์†๋„๋กœ ๋ฐœ์ „ํ•ด ์™”์Šต๋‹ˆ๋‹ค. ์ปดํ“จํ„ฐ ๋น„์ „์—์„œ ๊ฐ€์žฅ ์ค‘์š”ํ•œ ๋ฐ์ดํ„ฐ ์„ธํŠธ ์ค‘ ํ•˜๋‚˜๋Š” ์ด๋ฏธ์ง€ ๋ถ„๋ฅ˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ํ…Œ.. 2023. 3. 23.