์ฝ˜ํ…์ธ ๋กœ ๊ฑด๋„ˆ๋›ฐ๊ธฐ

YOLOv7: ํ›ˆ๋ จ ๊ฐ€๋Šฅํ•œ ๊ณต์งœ ๊ฐ€๋ฐฉ

YOLOv7์€ ์†๋„์™€ ์ •ํ™•๋„ ๋ชจ๋‘์—์„œ ์•Œ๋ ค์ง„ ๋ชจ๋“  ๋ฌผ์ฒด ๊ฐ์ง€๊ธฐ๋ฅผ ๋Šฅ๊ฐ€ํ•˜๋Š” 5 FPS ~ 160 FPS ๋ฒ”์œ„์˜ ์ตœ์ฒจ๋‹จ ์‹ค์‹œ๊ฐ„ ๋ฌผ์ฒด ๊ฐ์ง€๊ธฐ์ž…๋‹ˆ๋‹ค. GPU V100์—์„œ 30 FPS ์ด์ƒ์˜ ๋ชจ๋“  ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฒ€์ถœ๊ธฐ ์ค‘ ๊ฐ€์žฅ ๋†’์€ ์ •ํ™•๋„(56.8% AP)๋ฅผ ์ž๋ž‘ํ•ฉ๋‹ˆ๋‹ค. ๋˜ํ•œ YOLOv7์€ ์†๋„์™€ ์ •ํ™•๋„ ๋ฉด์—์„œ YOLOR, YOLOX, Scaled-YOLOv4, YOLOv5 ๋“ฑ๊ณผ ๊ฐ™์€ ๋‹ค๋ฅธ ๊ฐ์ฒด ๊ฒ€์ถœ๊ธฐ๋ณด๋‹ค ์„ฑ๋Šฅ์ด ๋›ฐ์–ด๋‚ฉ๋‹ˆ๋‹ค. ์ด ๋ชจ๋ธ์€ ๋‹ค๋ฅธ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋‚˜ ์‚ฌ์ „ ํ•™์Šต๋œ ๊ฐ€์ค‘์น˜๋ฅผ ์‚ฌ์šฉํ•˜์ง€ ์•Š๊ณ  ์ฒ˜์Œ๋ถ€ํ„ฐ MS COCO ๋ฐ์ดํ„ฐ ์„ธํŠธ์— ๋Œ€ํ•ด ํ•™์Šต๋ฉ๋‹ˆ๋‹ค. YOLOv7์˜ ์†Œ์Šค ์ฝ”๋“œ๋Š” GitHub์—์„œ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

YOLOv7๊ณผ SOTA ๋ฌผ์ฒด ๊ฐ์ง€๊ธฐ์˜ ๋น„๊ต

SOTA ๋ฌผ์ฒด ๊ฐ์ง€๊ธฐ ๋น„๊ต

YOLO ๋น„๊ต ํ‘œ์˜ ๊ฒฐ๊ณผ๋ฅผ ๋ณด๋ฉด ์ œ์•ˆํ•œ ๋ฐฉ๋ฒ•์ด ์†๋„์™€ ์ •ํ™•๋„ ๊ฐ„์˜ ๊ท ํ˜•์„ ์ข…ํ•ฉ์ ์œผ๋กœ ๊ฐ€์žฅ ์ž˜ ๋งž์ถ”๊ณ  ์žˆ๋‹ค๋Š” ๊ฒƒ์„ ์•Œ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. YOLOv7-tiny-SiLU์™€ YOLOv5-N(r6.1)์„ ๋น„๊ตํ•˜๋ฉด, ์šฐ๋ฆฌ ๋ฐฉ์‹์ด 127fps ๋” ๋น ๋ฅด๊ณ  AP์—์„œ 10.7% ๋” ์ •ํ™•ํ•ฉ๋‹ˆ๋‹ค. ๋˜ํ•œ YOLOv7์€ 161fps์˜ ํ”„๋ ˆ์ž„ ์†๋„์—์„œ 51.4%์˜ AP๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๋ฐ˜๋ฉด, ๋™์ผํ•œ AP๋ฅผ ์‚ฌ์šฉํ•˜๋Š” PPYOLOE-L์˜ ํ”„๋ ˆ์ž„ ์†๋„๋Š” 78fps์— ๋ถˆ๊ณผํ•ฉ๋‹ˆ๋‹ค. ํŒŒ๋ผ๋ฏธํ„ฐ ์‚ฌ์šฉ๋Ÿ‰ ์ธก๋ฉด์—์„œ๋Š” YOLOv7์ด PPYOLOE-L๋ณด๋‹ค 41% ๋” ์ ์Šต๋‹ˆ๋‹ค. ์ถ”๋ก  ์†๋„๊ฐ€ 114fps์ธ YOLOv7-X์™€ ์ถ”๋ก  ์†๋„๊ฐ€ 99fps์ธ YOLOv5-L(r6.1)์„ ๋น„๊ตํ•˜๋ฉด, YOLOv7-X๋Š” AP๋ฅผ 3.9% ํ–ฅ์ƒ์‹œํ‚ฌ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๋น„์Šทํ•œ ๊ทœ๋ชจ์˜ YOLOv5-X(r6.1)์™€ ๋น„๊ตํ•˜๋ฉด, YOLOv7-X์˜ ์ถ”๋ก  ์†๋„๊ฐ€ 31fps ๋” ๋น ๋ฆ…๋‹ˆ๋‹ค. ๋˜ํ•œ ํŒŒ๋ผ๋ฏธํ„ฐ์™€ ์—ฐ์‚ฐ๋Ÿ‰ ์ธก๋ฉด์—์„œ๋„ YOLOv7-X๋Š” YOLOv5-X(r6.1)์— ๋น„ํ•ด ํŒŒ๋ผ๋ฏธํ„ฐ๋Š” 22%, ์—ฐ์‚ฐ๋Ÿ‰์€ 8% ๊ฐ์†Œํ•˜์ง€๋งŒ AP๋Š” 2.2% ํ–ฅ์ƒ๋ฉ๋‹ˆ๋‹ค(์ถœ์ฒ˜).

๋ชจ๋ธ ๋งค๊ฐœ๋ณ€์ˆ˜
(M)
ํ”Œ๋กญ
(G)
ํฌ๊ธฐ
(ํ”ฝ์…€)
FPS APtest/ val
50-95
APtest
50
APtest
75
APtest
S
APtest
M
APtest
L
YOLOX-S 9.0M 26.8G 640 102 40.5% / 40.5% - - - - -
YOLOX-M 25.3M 73.8G 640 81 47.2% / 46.9% - - - - -
YOLOX-L 54.2M 155.6G 640 69 50.1% / 49.7% - - - - -
YOLOX-X 99.1M 281.9G 640 58 51.5% / 51.1% - - - - -
PPYOLOE-S 7.9M 17.4G 640 208 43.1% / 42.7% 60.5% 46.6% 23.2% 46.4% 56.9%
PPYOLOE-M 23.4M 49.9G 640 123 48.9% / 48.6% 66.5% 53.0% 28.6% 52.9% 63.8%
PPYOLOE-L 52.2M 110.1G 640 78 51.4% / 50.9% 68.9% 55.6% 31.4% 55.3% 66.1%
PPYOLOE-X 98.4M 206.6G 640 45 52.2% / 51.9% 69.9% 56.5% 33.3% 56.3% 66.4%
YOLOv5-N(r6.1) 1.9M 4.5G 640 159 - / 28.0% - - - - -
YOLOv5-S(r6.1) 7.2M 16.5G 640 156 - / 37.4% - - - - -
YOLOv5-M(r6.1) 21.2M 49.0G 640 122 - / 45.4% - - - - -
YOLOv5-L(r6.1) 46.5M 109.1G 640 99 - / 49.0% - - - - -
YOLOv5-X(r6.1) 86.7M 205.7G 640 83 - / 50.7% - - - - -
YOLOR-CSP 52.9M 120.4G 640 106 51.1% / 50.8% 69.6% 55.7% 31.7% 55.3% 64.7%
YOLOR-CSP-X 96.9M 226.8G 640 87 53.0% / 52.7% 71.4% 57.9% 33.7% 57.1% 66.8%
YOLOv7-tiny-SiLU 6.2M 13.8G 640 286 38.7% / 38.7% 56.7% 41.7% 18.8% 42.4% 51.9%
YOLOv7 36.9M 104.7G 640 161 51.4% / 51.2% 69.7% 55.9% 31.8% 55.5% 65.0%
YOLOv7-X 71.3M 189.9G 640 114 53.1% / 52.9% 71.2% 57.8% 33.8% 57.1% 67.4%
YOLOv5-N6(r6.1) 3.2M 18.4G 1280 123 - / 36.0% - - - - -
YOLOv5-S6(r6.1) 12.6M 67.2G 1280 122 - / 44.8% - - - - -
YOLOv5-M6(r6.1) 35.7M 200.0G 1280 90 - / 51.3% - - - - -
YOLOv5-L6(r6.1) 76.8M 445.6G 1280 63 - / 53.7% - - - - -
YOLOv5-X6(r6.1) 140.7M 839.2G 1280 38 - / 55.0% - - - - -
YOLOR-P6 37.2M 325.6G 1280 76 53.9% / 53.5% 71.4% 58.9% 36.1% 57.7% 65.6%
YOLOR-W6 79.8G 453.2G 1280 66 55.2% / 54.8% 72.7% 60.5% 37.7% 59.1% 67.1%
YOLOR-E6 115.8M 683.2G 1280 45 55.8% / 55.7% 73.4% 61.1% 38.4% 59.7% 67.7%
YOLOR-D6 151.7M 935.6G 1280 34 56.5% / 56.1% 74.1% 61.9% 38.9% 60.4% 68.7%
YOLOv7-W6 70.4M 360.0G 1280 84 54.9% / 54.6% 72.6% 60.1% 37.3% 58.7% 67.1%
YOLOv7-E6 97.2M 515.2G 1280 56 56.0% / 55.9% 73.5% 61.2% 38.0% 59.9% 68.4%
YOLOv7-D6 154.7M 806.8G 1280 44 56.6% / 56.3% 74.0% 61.8% 38.8% 60.1% 69.5%
YOLOv7-E6E 151.7M 843.2G 1280 36 56.8% / 56.8% 74.4% 62.1% 39.3% 60.5% 69.0%

๊ฐœ์š”

์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€๋Š” ๋‹ค์ค‘ ๊ฐ์ฒด ์ถ”์ , ์ž์œจ ์ฃผํ–‰, ๋กœ๋ด‡ ๊ณตํ•™, ์˜๋ฃŒ ์ด๋ฏธ์ง€ ๋ถ„์„ ๋“ฑ ๋งŽ์€ ์ปดํ“จํ„ฐ ๋น„์ „ ์‹œ์Šคํ…œ์—์„œ ์ค‘์š”ํ•œ ๊ตฌ์„ฑ ์š”์†Œ์ž…๋‹ˆ๋‹ค. ์ตœ๊ทผ ๋ช‡ ๋…„ ๋™์•ˆ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€ ๊ฐœ๋ฐœ์€ ํšจ์œจ์ ์ธ ์•„ํ‚คํ…์ฒ˜๋ฅผ ์„ค๊ณ„ํ•˜๊ณ  ๋‹ค์–‘ํ•œ CPU, GPU ๋ฐ ์‹ ๊ฒฝ ์ฒ˜๋ฆฌ ์žฅ์น˜(NPU)์˜ ์ถ”๋ก  ์†๋„๋ฅผ ๊ฐœ์„ ํ•˜๋Š” ๋ฐ ์ค‘์ ์„ ๋‘๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. YOLOv7์€ ์—ฃ์ง€๋ถ€ํ„ฐ ํด๋ผ์šฐ๋“œ๊นŒ์ง€ ๋ชจ๋ฐ”์ผ GPU ๋ฐ GPU ๋””๋ฐ”์ด์Šค๋ฅผ ๋ชจ๋‘ ์ง€์›ํ•ฉ๋‹ˆ๋‹ค.

์•„ํ‚คํ…์ฒ˜ ์ตœ์ ํ™”์— ์ดˆ์ ์„ ๋งž์ถ˜ ๊ธฐ์กด์˜ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€๊ธฐ์™€ ๋‹ฌ๋ฆฌ, YOLOv7์€ ํ•™์Šต ๊ณผ์ •์˜ ์ตœ์ ํ™”์— ์ค‘์ ์„ ๋‘ก๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” 'ํ›ˆ๋ จ ๊ฐ€๋Šฅํ•œ ๊ณต์งœ ๊ฐ€๋ฐฉ'์ด๋ผ๋Š” ๊ฐœ๋…์œผ๋กœ ์•Œ๋ ค์ง„ ์ถ”๋ก  ๋น„์šฉ์„ ๋Š˜๋ฆฌ์ง€ ์•Š๊ณ  ๊ฐ์ฒด ๊ฐ์ง€์˜ ์ •ํ™•๋„๋ฅผ ํ–ฅ์ƒํ•˜๋„๋ก ์„ค๊ณ„๋œ ๋ชจ๋“ˆ๊ณผ ์ตœ์ ํ™” ๋ฐฉ๋ฒ•์ด ํฌํ•จ๋ฉ๋‹ˆ๋‹ค.

์ฃผ์š” ๊ธฐ๋Šฅ

YOLOv7์€ ๋ช‡ ๊ฐ€์ง€ ์ฃผ์š” ๊ธฐ๋Šฅ์„ ์†Œ๊ฐœํ•ฉ๋‹ˆ๋‹ค:

  1. ๋ชจ๋ธ ์žฌํŒŒ๋ผ๋ฏธํ„ฐํ™”: YOLOv7์€ ๊ฒฝ์‚ฌ ์ „ํŒŒ ๊ฒฝ๋กœ๋ผ๋Š” ๊ฐœ๋…์œผ๋กœ ๋‹ค์–‘ํ•œ ๋„คํŠธ์›Œํฌ์˜ ๋ ˆ์ด์–ด์— ์ ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ์ „๋žต์ธ ๊ณ„ํš๋œ ์žฌํŒŒ๋ผ๋ฏธํ„ฐํ™” ๋ชจ๋ธ์„ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค.

  2. ๋™์  ๋ ˆ์ด๋ธ” ํ• ๋‹น: ์—ฌ๋Ÿฌ ์ถœ๋ ฅ ๋ ˆ์ด์–ด๊ฐ€ ์žˆ๋Š” ๋ชจ๋ธ์„ ํ•™์Šตํ•  ๋•Œ ์ƒˆ๋กœ์šด ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•ฉ๋‹ˆ๋‹ค: "์„œ๋กœ ๋‹ค๋ฅธ ๋ถ„๊ธฐ์˜ ์ถœ๋ ฅ์— ๋™์  ๋ชฉํ‘œ๋ฅผ ์–ด๋–ป๊ฒŒ ํ• ๋‹นํ•  ๊ฒƒ์ธ๊ฐ€?"๋ผ๋Š” ๋ฌธ์ œ์ž…๋‹ˆ๋‹ค. ์ด ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด YOLOv7์€ ๊ฑฐ์น ๊ณ  ์„ธ๋ฐ€ํ•œ ๋ฆฌ๋“œ ๊ฐ€์ด๋“œ ๋ผ๋ฒจ ํ• ๋‹น์ด๋ผ๋Š” ์ƒˆ๋กœ์šด ๋ผ๋ฒจ ํ• ๋‹น ๋ฐฉ๋ฒ•์„ ๋„์ž…ํ–ˆ์Šต๋‹ˆ๋‹ค.

  3. ํ™•์žฅ ๋ฐ ๋ณตํ•ฉ ์Šค์ผ€์ผ๋ง: YOLOv7์€ ํŒŒ๋ผ๋ฏธํ„ฐ์™€ ๊ณ„์‚ฐ์„ ํšจ๊ณผ์ ์œผ๋กœ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€๊ธฐ์˜ 'ํ™•์žฅ' ๋ฐ '๋ณตํ•ฉ ์Šค์ผ€์ผ๋ง' ๋ฐฉ๋ฒ•์„ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค.

  4. ํšจ์œจ์„ฑ: YOLOv7์ด ์ œ์•ˆํ•˜๋Š” ๋ฐฉ๋ฒ•์€ ์ตœ์ฒจ๋‹จ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€๊ธฐ์˜ ์•ฝ 40%์˜ ํŒŒ๋ผ๋ฏธํ„ฐ์™€ 50%์˜ ๊ณ„์‚ฐ์„ ํšจ๊ณผ์ ์œผ๋กœ ์ค„์ผ ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ถ”๋ก  ์†๋„๊ฐ€ ๋” ๋น ๋ฅด๊ณ  ๊ฐ์ง€ ์ •ํ™•๋„๊ฐ€ ๋” ๋†’์Šต๋‹ˆ๋‹ค.

์‚ฌ์šฉ ์˜ˆ

์ด ๊ธ€์„ ์ž‘์„ฑํ•˜๋Š” ์‹œ์ ์„ ๊ธฐ์ค€์œผ๋กœ Ultralytics ์€ ํ˜„์žฌ YOLOv7 ๋ชจ๋ธ์„ ์ง€์›ํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ YOLOv7 ์‚ฌ์šฉ์— ๊ด€์‹ฌ์ด ์žˆ๋Š” ์‚ฌ์šฉ์ž๋Š” ์„ค์น˜ ๋ฐ ์‚ฌ์šฉ ์ง€์นจ์„ YOLOv7 GitHub ๋ฆฌํฌ์ง€ํ† ๋ฆฌ์—์„œ ์ง์ ‘ ์ฐธ์กฐํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.

๋‹ค์Œ์€ YOLOv7์„ ์‚ฌ์šฉํ•˜๊ธฐ ์œ„ํ•ด ์ทจํ•  ์ˆ˜ ์žˆ๋Š” ์ผ๋ฐ˜์ ์ธ ๋‹จ๊ณ„์— ๋Œ€ํ•œ ๊ฐ„๋žตํ•œ ๊ฐœ์š”์ž…๋‹ˆ๋‹ค:

  1. YOLOv7 GitHub ๋ฆฌํฌ์ง€ํ† ๋ฆฌ (https://github.com/WongKinYiu/yolov7)๋ฅผ ๋ฐฉ๋ฌธํ•˜์„ธ์š”.

  2. README ํŒŒ์ผ์— ์ œ๊ณต๋œ ์ง€์นจ์— ๋”ฐ๋ผ ์„ค์น˜ํ•ฉ๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์ผ๋ฐ˜์ ์œผ๋กœ ๋ฆฌํฌ์ง€ํ† ๋ฆฌ ๋ณต์ œ, ํ•„์š”ํ•œ ์ข…์†์„ฑ ์„ค์น˜, ํ•„์š”ํ•œ ํ™˜๊ฒฝ ๋ณ€์ˆ˜ ์„ค์ •์ด ํฌํ•จ๋ฉ๋‹ˆ๋‹ค.

  3. ์„ค์น˜๊ฐ€ ์™„๋ฃŒ๋˜๋ฉด ๋ฆฌํฌ์ง€ํ† ๋ฆฌ์— ์ œ๊ณต๋œ ์‚ฌ์šฉ ์ง€์นจ์— ๋”ฐ๋ผ ๋ชจ๋ธ์„ ํ›ˆ๋ จํ•˜๊ณ  ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์ผ๋ฐ˜์ ์œผ๋กœ ๋ฐ์ดํ„ฐ ์„ธํŠธ ์ค€๋น„, ๋ชจ๋ธ ๋งค๊ฐœ๋ณ€์ˆ˜ ๊ตฌ์„ฑ, ๋ชจ๋ธ ํ•™์Šต, ํ•™์Šต๋œ ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•˜์—ฌ ๊ฐ์ฒด ๊ฐ์ง€๋ฅผ ์ˆ˜ํ–‰ํ•˜๋Š” ๊ณผ์ •์ด ํฌํ•จ๋ฉ๋‹ˆ๋‹ค.

๊ตฌ์ฒด์ ์ธ ๋‹จ๊ณ„๋Š” ํŠน์ • ์‚ฌ์šฉ ์‚ฌ๋ก€์™€ YOLOv7 ๋ฆฌํฌ์ง€ํ† ๋ฆฌ์˜ ํ˜„์žฌ ์ƒํƒœ์— ๋”ฐ๋ผ ๋‹ฌ๋ผ์งˆ ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ์— ์œ ์˜ํ•˜์„ธ์š”. ๋”ฐ๋ผ์„œ YOLOv7 GitHub ๋ฆฌํฌ์ง€ํ† ๋ฆฌ์— ์ œ๊ณต๋œ ์ง€์นจ์„ ์ง์ ‘ ์ฐธ์กฐํ•˜๋Š” ๊ฒƒ์ด ์ข‹์Šต๋‹ˆ๋‹ค.

์ด๋กœ ์ธํ•ด ๋ถˆํŽธ์„ ๋“œ๋ ค ์ฃ„์†กํ•˜๋ฉฐ, YOLOv7์— ๋Œ€ํ•œ ์ง€์›์ด ๊ตฌํ˜„๋˜๋Š” ๋Œ€๋กœ Ultralytics ์— ์‚ฌ์šฉ ์˜ˆ์‹œ๊ฐ€ ํฌํ•จ๋œ ๋ฌธ์„œ๋ฅผ ์—…๋ฐ์ดํŠธํ•˜๋„๋ก ๋…ธ๋ ฅํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.

์ธ์šฉ ๋ฐ ๊ฐ์‚ฌ

์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€ ๋ถ„์•ผ์—์„œ ํฌ๊ฒŒ ๊ธฐ์—ฌํ•œ YOLOv7 ์ž‘์„ฑ์ž์—๊ฒŒ ๊ฐ์‚ฌ์˜ ๋ง์”€์„ ์ „ํ•ฉ๋‹ˆ๋‹ค:

@article{wang2022yolov7,
  title={YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors},
  author={Wang, Chien-Yao and Bochkovskiy, Alexey and Liao, Hong-Yuan Mark},
  journal={arXiv preprint arXiv:2207.02696},
  year={2022}
}

YOLOv7 ๋…ผ๋ฌธ ์›๋ณธ์€ arXiv์—์„œ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ €์ž๋“ค์€ ์ž์‹ ์˜ ์ž‘์—…์„ ๊ณต๊ฐœํ–ˆ์œผ๋ฉฐ, ์ฝ”๋“œ๋ฒ ์ด์Šค๋Š” GitHub์—์„œ ์•ก์„ธ์Šคํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด ๋ถ„์•ผ๋ฅผ ๋ฐœ์ „์‹œํ‚ค๊ณ  ๋” ๋งŽ์€ ์ปค๋ฎค๋‹ˆํ‹ฐ๊ฐ€ ์ž์‹ ์˜ ์—ฐ๊ตฌ์— ์ ‘๊ทผํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•œ ์ €์ž๋“ค์˜ ๋…ธ๋ ฅ์— ๊ฐ์‚ฌ๋“œ๋ฆฝ๋‹ˆ๋‹ค.

์ž์ฃผ ๋ฌป๋Š” ์งˆ๋ฌธ

YOLOv7์ด๋ž€ ๋ฌด์—‡์ด๋ฉฐ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€์˜ ํš๊ธฐ์ ์ธ ๊ธฐ์ˆ ๋กœ ๊ฐ„์ฃผ๋˜๋Š” ์ด์œ ๋Š” ๋ฌด์—‡์ธ๊ฐ€์š”?

YOLOv7์€ ๋น„๊ตํ•  ์ˆ˜ ์—†๋Š” ์†๋„์™€ ์ •ํ™•๋„๋ฅผ ์ž๋ž‘ํ•˜๋Š” ์ตœ์ฒจ๋‹จ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. ๋งค๊ฐœ๋ณ€์ˆ˜ ์‚ฌ์šฉ๊ณผ ์ถ”๋ก  ์†๋„ ๋ชจ๋‘์—์„œ YOLOX, YOLOv5, PPYOLOE์™€ ๊ฐ™์€ ๋‹ค๋ฅธ ๋ชจ๋ธ์„ ๋Šฅ๊ฐ€ํ•ฉ๋‹ˆ๋‹ค. ๋ชจ๋ธ ์žฌํŒŒ๋ผ๋ฏธํ„ฐํ™”์™€ ๋™์  ๋ผ๋ฒจ ํ• ๋‹น์„ ํ†ตํ•ด ์ถ”๋ก  ๋น„์šฉ์„ ์ฆ๊ฐ€์‹œํ‚ค์ง€ ์•Š๊ณ  ์„ฑ๋Šฅ์„ ์ตœ์ ํ™”ํ•˜๋Š” ๊ฒƒ์ด YOLOv7์˜ ์ฐจ๋ณ„ํ™”๋œ ํŠน์ง•์ž…๋‹ˆ๋‹ค. ์•„ํ‚คํ…์ฒ˜์™€ ๋‹ค๋ฅธ ์ตœ์‹  ๊ฐ์ฒด ๊ฐ์ง€๊ธฐ์™€์˜ ๋น„๊ต ๋ฉ”ํŠธ๋ฆญ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ๊ธฐ์ˆ  ์ •๋ณด๋Š” YOLOv7 ๋ฐฑ์„œ๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”.

YOLOv7์€ ์ด์ „ YOLO ๋ชจ๋ธ์ธ YOLOv4์™€ YOLOv5 ๋ณด๋‹ค ์–ด๋–ป๊ฒŒ ๊ฐœ์„ ๋˜์—ˆ๋‚˜์š”?

YOLOv7์€ ๋ชจ๋ธ ์žฌํŒŒ๋ผ๋ฏธํ„ฐํ™” ๋ฐ ๋™์  ๋ผ๋ฒจ ํ• ๋‹น ๋“ฑ ๋ช‡ ๊ฐ€์ง€ ํ˜์‹  ๊ธฐ๋Šฅ์„ ๋„์ž…ํ•˜์—ฌ ํ•™์Šต ํ”„๋กœ์„ธ์Šค๋ฅผ ๊ฐœ์„ ํ•˜๊ณ  ์ถ”๋ก  ์ •ํ™•๋„๋ฅผ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค. YOLOv5 ๊ณผ ๋น„๊ตํ–ˆ์„ ๋•Œ YOLOv7์€ ์†๋„์™€ ์ •ํ™•๋„๊ฐ€ ํฌ๊ฒŒ ํ–ฅ์ƒ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, YOLOv7-X๋Š” YOLOv5-X์— ๋น„ํ•ด ์ •ํ™•๋„๋Š” 2.2% ํ–ฅ์ƒ๋˜๊ณ  ํŒŒ๋ผ๋ฏธํ„ฐ๋Š” 22% ๊ฐ์†Œํ•ฉ๋‹ˆ๋‹ค. ์ž์„ธํ•œ ๋น„๊ต๋Š” ์„ฑ๋Šฅ ํ‘œ YOLOv7๊ณผ SOTA ๊ฐ์ฒด ๊ฐ์ง€๊ธฐ ๋น„๊ต์—์„œ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

Ultralytics ๋„๊ตฌ ๋ฐ ํ”Œ๋žซํผ์—์„œ YOLOv7์„ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋‚˜์š”?

ํ˜„์žฌ Ultralytics ๋Š” ๋„๊ตฌ ๋ฐ ํ”Œ๋žซํผ์—์„œ YOLOv7์„ ์ง์ ‘ ์ง€์›ํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค. YOLOv7 ์‚ฌ์šฉ์— ๊ด€์‹ฌ์ด ์žˆ๋Š” ์‚ฌ์šฉ์ž๋Š” YOLOv7 GitHub ๋ฆฌํฌ์ง€ํ† ๋ฆฌ์— ์ œ๊ณต๋œ ์„ค์น˜ ๋ฐ ์‚ฌ์šฉ ์ง€์นจ์„ ๋”ฐ๋ผ์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๋‹ค๋ฅธ ์ตœ์‹  ๋ชจ๋ธ์˜ ๊ฒฝ์šฐ Ultralytics ํ—ˆ๋ธŒ์™€ ๊ฐ™์€ Ultralytics ๋„๊ตฌ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ํƒ์ƒ‰ํ•˜๊ณ  ๊ต์œกํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

์‚ฌ์šฉ์ž ์ง€์ • ๊ฐ์ฒด ๊ฐ์ง€ ํ”„๋กœ์ ํŠธ๋ฅผ ์œ„ํ•ด YOLOv7์„ ์„ค์น˜ํ•˜๊ณ  ์‹คํ–‰ํ•˜๋ ค๋ฉด ์–ด๋–ป๊ฒŒ ํ•ด์•ผ ํ•˜๋‚˜์š”?

YOLOv7์„ ์„ค์น˜ํ•˜๊ณ  ์‹คํ–‰ํ•˜๋ ค๋ฉด ๋‹ค์Œ ๋‹จ๊ณ„๋ฅผ ๋”ฐ๋ฅด์„ธ์š”:

  1. YOLOv7 ๋ฆฌํฌ์ง€ํ† ๋ฆฌ๋ฅผ ๋ณต์ œํ•ฉ๋‹ˆ๋‹ค:
    git clone https://github.com/WongKinYiu/yolov7
    
  2. ๋ณต์ œ๋œ ๋””๋ ‰ํ† ๋ฆฌ๋กœ ์ด๋™ํ•˜์—ฌ ์ข…์† ์š”์†Œ๋ฅผ ์„ค์น˜ํ•ฉ๋‹ˆ๋‹ค:
    cd yolov7
    pip install -r requirements.txt
    
  3. ์ €์žฅ์†Œ์— ์ œ๊ณต๋œ ์‚ฌ์šฉ ์ง€์นจ์— ๋”ฐ๋ผ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ์ค€๋น„ํ•˜๊ณ  ๋ชจ๋ธ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ๊ตฌ์„ฑํ•ฉ๋‹ˆ๋‹ค. ์ž์„ธํ•œ ์ง€์นจ์€ YOLOv7 GitHub ๋ฆฌํฌ์ง€ํ† ๋ฆฌ์—์„œ ์ตœ์‹  ์ •๋ณด ๋ฐ ์—…๋ฐ์ดํŠธ๋ฅผ ํ™•์ธํ•˜์„ธ์š”.

YOLOv7์— ๋„์ž…๋œ ์ฃผ์š” ๊ธฐ๋Šฅ ๋ฐ ์ตœ์ ํ™”๋Š” ๋ฌด์—‡์ธ๊ฐ€์š”?

YOLOv7์€ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€๋ฅผ ํ˜์‹ ์ ์œผ๋กœ ๊ฐœ์„ ํ•˜๋Š” ๋ช‡ ๊ฐ€์ง€ ์ฃผ์š” ๊ธฐ๋Šฅ์„ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค:

  • ๋ชจ๋ธ ์žฌํŒŒ๋ผ๋ฏธํ„ฐํ™”: ๊ทธ๋ผ๋ฐ์ด์…˜ ์ „ํŒŒ ๊ฒฝ๋กœ๋ฅผ ์ตœ์ ํ™”ํ•˜์—ฌ ๋ชจ๋ธ์˜ ์„ฑ๋Šฅ์„ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.
  • ๋™์  ๋ ˆ์ด๋ธ” ํ• ๋‹น: ๊ฑฐ์น ๊ณ  ๋ฏธ์„ธํ•œ ๋ฆฌ๋“œ ๊ฐ€์ด๋“œ ๋ฐฉ๋ฒ•์„ ์‚ฌ์šฉํ•˜์—ฌ ์—ฌ๋Ÿฌ ์ง€์ ์— ๊ฑธ์ณ ์ถœ๋ ฅ๋ฌผ์— ๋Œ€ํ•œ ๋™์  ๋Œ€์ƒ์„ ํ• ๋‹นํ•˜์—ฌ ์ •ํ™•๋„๋ฅผ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.
  • ํ™•์žฅ ๋ฐ ๋ณตํ•ฉ ์Šค์ผ€์ผ๋ง: ํŒŒ๋ผ๋ฏธํ„ฐ์™€ ๊ณ„์‚ฐ์„ ํšจ์œจ์ ์œผ๋กœ ํ™œ์šฉํ•˜์—ฌ ๋‹ค์–‘ํ•œ ์‹ค์‹œ๊ฐ„ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์— ๋งž๊ฒŒ ๋ชจ๋ธ์„ ํ™•์žฅํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
  • ํšจ์œจ์„ฑ: ๋‹ค๋ฅธ ์ตœ์‹  ๋ชจ๋ธ์— ๋น„ํ•ด ๋งค๊ฐœ๋ณ€์ˆ˜ ์ˆ˜๋Š” 40%, ๊ณ„์‚ฐ์€ 50% ์ค„์ด๋ฉด์„œ ์ถ”๋ก  ์†๋„๋Š” ๋” ๋นจ๋ผ์ง‘๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ๊ธฐ๋Šฅ์— ๋Œ€ํ•œ ์ž์„ธํ•œ ๋‚ด์šฉ์€ YOLOv7 ๊ฐœ์š” ์„น์…˜์„ ์ฐธ์กฐํ•˜์„ธ์š”.
๐Ÿ“…1 ๋…„ ์ „ ์ƒ์„ฑ๋จ โœ๏ธ 1๊ฐœ์›” ์ „ ์—…๋ฐ์ดํŠธ๋จ

๋Œ“๊ธ€