์ฝ˜ํ…์ธ ๋กœ ๊ฑด๋„ˆ๋›ฐ๊ธฐ

YOLOv10: ์‹ค์‹œ๊ฐ„ ์—”๋“œํˆฌ์—”๋“œ ๊ฐ์ฒด ๊ฐ์ง€

ํŒจํ‚ค์ง€๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋œ YOLOv10์€ Ultralytics Python ํŒจํ‚ค์ง€๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋œ YOLOv10์€ ์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€์— ๋Œ€ํ•œ ์ƒˆ๋กœ์šด ์ ‘๊ทผ ๋ฐฉ์‹์„ ๋„์ž…ํ•˜์—ฌ ์ด์ „ YOLO ๋ฒ„์ „์—์„œ ๋ฐœ๊ฒฌ๋œ ํ›„์ฒ˜๋ฆฌ ๋ฐ ๋ชจ๋ธ ์•„ํ‚คํ…์ฒ˜์˜ ๊ฒฐํ•จ์„ ๋ชจ๋‘ ํ•ด๊ฒฐํ–ˆ์Šต๋‹ˆ๋‹ค. ๋น„์ตœ๋Œ€ ์–ต์ œ(NMS)๋ฅผ ์ œ๊ฑฐํ•˜๊ณ  ๋‹ค์–‘ํ•œ ๋ชจ๋ธ ๊ตฌ์„ฑ ์š”์†Œ๋ฅผ ์ตœ์ ํ™”ํ•จ์œผ๋กœ์จ YOLOv10์€ ๊ณ„์‚ฐ ์˜ค๋ฒ„ํ—ค๋“œ๋ฅผ ํฌ๊ฒŒ ์ค„์ด๋ฉด์„œ ์ตœ์ฒจ๋‹จ ์„ฑ๋Šฅ์„ ๋‹ฌ์„ฑํ•ฉ๋‹ˆ๋‹ค. ๊ด‘๋ฒ”์œ„ํ•œ ์‹คํ—˜์„ ํ†ตํ•ด ์—ฌ๋Ÿฌ ๋ชจ๋ธ ๊ทœ๋ชจ์— ๊ฑธ์ณ ์ •ํ™•๋„์™€ ์ง€์—ฐ ์‹œ๊ฐ„ ๊ฐ„์˜ ์šฐ์ˆ˜ํ•œ ์ ˆ์ถฉ์ ์„ ์ž…์ฆํ–ˆ์Šต๋‹ˆ๋‹ค.

NMS ์—†๋Š” ๊ต์œก์„ ์œ„ํ•œ YOLOv10 ์ผ๊ด€๋œ ์ด์ค‘ ํ• ๋‹น

๊ฐœ์š”

์‹ค์‹œ๊ฐ„ ๊ฐ์ฒด ๊ฐ์ง€๋Š” ์งง์€ ์ง€์—ฐ ์‹œ๊ฐ„์œผ๋กœ ์ด๋ฏธ์ง€์—์„œ ๊ฐ์ฒด ๋ฒ”์ฃผ์™€ ์œ„์น˜๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ์˜ˆ์ธกํ•˜๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•ฉ๋‹ˆ๋‹ค. YOLO ์‹œ๋ฆฌ์ฆˆ๋Š” ์„ฑ๋Šฅ๊ณผ ํšจ์œจ์„ฑ ์‚ฌ์ด์˜ ๊ท ํ˜•์œผ๋กœ ์ธํ•ด ์ด ์—ฐ๊ตฌ์˜ ์„ ๋‘์— ์„œ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ NMS์— ๋Œ€ํ•œ ์˜์กด๋„์™€ ์•„ํ‚คํ…์ฒ˜์˜ ๋น„ํšจ์œจ์„ฑ์ด ์ตœ์ ์˜ ์„ฑ๋Šฅ์„ ์ €ํ•ดํ•ด ์™”์Šต๋‹ˆ๋‹ค. YOLOv10์€ NMS ์—†๋Š” ํ›ˆ๋ จ์„ ์œ„ํ•œ ์ผ๊ด€๋œ ์ด์ค‘ ๊ณผ์ œ์™€ ์ „์ฒด์ ์ธ ํšจ์œจ์„ฑ-์ •ํ™•๋„ ์ค‘์‹ฌ์˜ ๋ชจ๋ธ ์„ค๊ณ„ ์ „๋žต์„ ๋„์ž…ํ•˜์—ฌ ์ด๋Ÿฌํ•œ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•ฉ๋‹ˆ๋‹ค.

์•„ํ‚คํ…์ฒ˜

YOLOv10์˜ ์•„ํ‚คํ…์ฒ˜๋Š” ์ด์ „ YOLO ๋ชจ๋ธ์˜ ๊ฐ•์ ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๋ช‡ ๊ฐ€์ง€ ์ฃผ์š” ํ˜์‹ ์„ ๋„์ž…ํ–ˆ์Šต๋‹ˆ๋‹ค. ๋ชจ๋ธ ์•„ํ‚คํ…์ฒ˜๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๊ตฌ์„ฑ ์š”์†Œ๋กœ ์ด๋ฃจ์–ด์ ธ ์žˆ์Šต๋‹ˆ๋‹ค:

  1. ๋ฐฑ๋ณธ: ํŠน์ง• ์ถ”์ถœ์„ ๋‹ด๋‹นํ•˜๋Š” YOLOv10์˜ ๋ฐฑ๋ณธ์€ ํ–ฅ์ƒ๋œ ๋ฒ„์ „์˜ CSPNet(๊ต์ฐจ ๋‹จ๊ณ„ ๋ถ€๋ถ„ ๋„คํŠธ์›Œํฌ)์„ ์‚ฌ์šฉํ•˜์—ฌ ๊ทธ๋ผ๋ฐ์ด์…˜ ํ๋ฆ„์„ ๊ฐœ์„ ํ•˜๊ณ  ๊ณ„์‚ฐ ์ค‘๋ณต์„ฑ์„ ์ค„์ž…๋‹ˆ๋‹ค.
  2. Neck: The neck is designed to aggregate features from different scales and passes them to the head. It includes PAN (Path Aggregation Network) layers for effective multi-scale feature fusion.
  3. ์ผ๋Œ€๋‹ค ํ—ค๋“œ: ํ›ˆ๋ จ ์ค‘์— ๊ฐ์ฒด๋‹น ์—ฌ๋Ÿฌ ์˜ˆ์ธก์„ ์ƒ์„ฑํ•˜์—ฌ ํ’๋ถ€ํ•œ ๊ฐ๋… ์‹ ํ˜ธ๋ฅผ ์ œ๊ณตํ•˜๊ณ  ํ•™์Šต ์ •ํ™•๋„๋ฅผ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.
  4. ์ผ๋Œ€์ผ ํ—ค๋“œ: ์ถ”๋ก  ์ค‘์— ๊ฐ์ฒด๋‹น ํ•˜๋‚˜์˜ ์ตœ์  ์˜ˆ์ธก์„ ์ƒ์„ฑํ•˜์—ฌ NMS๊ฐ€ ํ•„์š”ํ•˜์ง€ ์•Š์œผ๋ฏ€๋กœ ๋Œ€๊ธฐ ์‹œ๊ฐ„์„ ์ค„์ด๊ณ  ํšจ์œจ์„ฑ์„ ๊ฐœ์„ ํ•ฉ๋‹ˆ๋‹ค.

์ฃผ์š” ๊ธฐ๋Šฅ

  1. NMS ์—†๋Š” ๊ต์œก: ์ผ๊ด€๋œ ์ด์ค‘ ํ• ๋‹น์„ ํ™œ์šฉํ•˜์—ฌ NMS์˜ ํ•„์š”์„ฑ์„ ์—†์• ๊ณ  ์ถ”๋ก  ๋Œ€๊ธฐ ์‹œ๊ฐ„์„ ์ค„์ž…๋‹ˆ๋‹ค.
  2. ์ „์ฒด๋ก ์  ๋ชจ๋ธ ์„ค๊ณ„: ๊ฒฝ๋Ÿ‰ ๋ถ„๋ฅ˜ ํ—ค๋“œ, ๊ณต๊ฐ„ ์ฑ„๋„ ๋ถ„๋ฆฌํ˜• ๋‹ค์šด ์ƒ˜ํ”Œ๋ง, ์ˆœ์œ„ ๊ฐ€์ด๋“œ ๋ธ”๋ก ์„ค๊ณ„ ๋“ฑ ํšจ์œจ์„ฑ๊ณผ ์ •ํ™•๋„ ์ธก๋ฉด์—์„œ ๋‹ค์–‘ํ•œ ๊ตฌ์„ฑ ์š”์†Œ๋ฅผ ์ข…ํ•ฉ์ ์œผ๋กœ ์ตœ์ ํ™”ํ•ฉ๋‹ˆ๋‹ค.
  3. ํ–ฅ์ƒ๋œ ๋ชจ๋ธ ๊ธฐ๋Šฅ: ๋Œ€๊ทœ๋ชจ ์ปค๋„ ์ปจ๋ณผ๋ฃจ์…˜๊ณผ ๋ถ€๋ถ„์ ์ธ ์…€ํ”„ ์–ดํ…์…˜ ๋ชจ๋“ˆ์„ ํ†ตํ•ฉํ•˜์—ฌ ํฐ ๊ณ„์‚ฐ ๋น„์šฉ ์—†์ด ์„ฑ๋Šฅ์„ ๊ฐœ์„ ํ•ฉ๋‹ˆ๋‹ค.

๋ชจ๋ธ ๋ณ€ํ˜•

YOLOv10์€ ๋‹ค์–‘ํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ์š”๊ตฌ ์‚ฌํ•ญ์„ ์ถฉ์กฑํ•  ์ˆ˜ ์žˆ๋„๋ก ๋‹ค์–‘ํ•œ ๋ชจ๋ธ ์Šค์ผ€์ผ๋กœ ์ œ๊ณต๋ฉ๋‹ˆ๋‹ค:

  • YOLOv10-N: ๋ฆฌ์†Œ์Šค๊ฐ€ ๊ทน๋„๋กœ ์ œํ•œ๋œ ํ™˜๊ฒฝ์„ ์œ„ํ•œ ๋‚˜๋…ธ ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.
  • YOLOv10-S: ์†๋„์™€ ์ •ํ™•์„ฑ์˜ ๊ท ํ˜•์„ ๋งž์ถ˜ ์†Œํ˜• ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.
  • YOLOv10-M: ์ผ๋ฐ˜ ์šฉ๋„์˜ ์ค‘๊ฐ„ ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.
  • YOLOv10-B: ์ •ํ™•๋„๋ฅผ ๋†’์ด๊ธฐ ์œ„ํ•ด ํญ์„ ๋Š˜๋ฆฐ ๊ท ํ˜• ์žกํžŒ ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.
  • YOLOv10-L: ๊ณ„์‚ฐ ๋ฆฌ์†Œ์Šค๊ฐ€ ์ฆ๊ฐ€ํ•˜์ง€๋งŒ ์ •ํ™•๋„๊ฐ€ ๋” ๋†’์€ ๋Œ€ํ˜• ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.
  • YOLOv10-X: ์ •ํ™•๋„์™€ ์„ฑ๋Šฅ์„ ๊ทน๋Œ€ํ™”ํ•˜๋Š” ์ดˆ๋Œ€ํ˜• ๋ฒ„์ „์ž…๋‹ˆ๋‹ค.

์„ฑ๋Šฅ

YOLOv10์€ ์ •ํ™•๋„์™€ ํšจ์œจ์„ฑ ์ธก๋ฉด์—์„œ ์ด์ „ YOLO ๋ฒ„์ „ ๋ฐ ๊ธฐํƒ€ ์ตœ์‹  ๋ชจ๋ธ๋ณด๋‹ค ์„ฑ๋Šฅ์ด ๋›ฐ์–ด๋‚ฉ๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, YOLOv10-S๋Š” COCO ๋ฐ์ดํ„ฐ ์„ธํŠธ์—์„œ ๋น„์Šทํ•œ AP๋ฅผ ์‚ฌ์šฉํ•˜๋Š” RT-DETR-R18๋ณด๋‹ค 1.8๋ฐฐ ๋น ๋ฅด๋ฉฐ, YOLOv10-B๋Š” ๋™์ผํ•œ ์„ฑ๋Šฅ์˜ YOLOv9-C๋ณด๋‹ค ์ง€์—ฐ ์‹œ๊ฐ„์ด 46% ์งง๊ณ  ๋งค๊ฐœ ๋ณ€์ˆ˜๊ฐ€ 25% ๋” ์ ์Šต๋‹ˆ๋‹ค.

๋ชจ๋ธ ์ž…๋ ฅ ํฌ๊ธฐ APval ํ”Œ๋กญ(G) ์ง€์—ฐ ์‹œ๊ฐ„(ms)
YOLOv10-N 640 38.5 6.7 1.84
YOLOv10-S 640 46.3 21.6 2.49
YOLOv10-M 640 51.1 59.1 4.74
YOLOv10-B 640 52.5 92.0 5.74
YOLOv10-L 640 53.2 120.3 7.28
YOLOv10-X 640 54.4 160.4 10.70

์ง€์—ฐ ์‹œ๊ฐ„์€ T4 GPU์—์„œ TensorRT FP16์œผ๋กœ ์ธก์ •ํ–ˆ์Šต๋‹ˆ๋‹ค.

๋ฐฉ๋ฒ•๋ก 

NMS ์—†๋Š” ๊ต์œก์„ ์œ„ํ•œ ์ผ๊ด€๋œ ์ด์ค‘ ํ• ๋‹น

YOLOv10์€ ํ’๋ถ€ํ•œ ๊ฐ๋…๊ณผ ํšจ์œจ์ ์ธ ์—”๋“œํˆฌ์—”๋“œ ๋ฐฐํฌ๋ฅผ ๋ณด์žฅํ•˜๊ธฐ ์œ„ํ•ด ํ›ˆ๋ จ ์ค‘์— ์ผ๋Œ€๋‹ค ๋ฐ ์ผ๋Œ€์ผ ์ „๋žต์„ ๊ฒฐํ•ฉํ•œ ์ด์ค‘ ๋ ˆ์ด๋ธ” ํ• ๋‹น์„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค. ์ผ๊ด€๋œ ๋งค์นญ ๋ฉ”ํŠธ๋ฆญ์€ ๋‘ ์ „๋žต ๊ฐ„์˜ ๊ฐ๋…์„ ์กฐ์ •ํ•˜์—ฌ ์ถ”๋ก  ์ค‘ ์˜ˆ์ธก์˜ ํ’ˆ์งˆ์„ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.

์ „์ฒด์ ์ธ ํšจ์œจ์„ฑ-์ •ํ™•์„ฑ ์ค‘์‹ฌ์˜ ๋ชจ๋ธ ์„ค๊ณ„

ํšจ์œจ์„ฑ ํ–ฅ์ƒ

  1. ๊ฒฝ๋Ÿ‰ ๋ถ„๋ฅ˜ ํ—ค๋“œ: ๊นŠ์ด๋ณ„๋กœ ๋ถ„๋ฆฌ ๊ฐ€๋Šฅํ•œ ์ปจ๋ณผ๋ฃจ์…˜์„ ์‚ฌ์šฉํ•˜์—ฌ ๋ถ„๋ฅ˜ ํ—ค๋“œ์˜ ๊ณ„์‚ฐ ์˜ค๋ฒ„ํ—ค๋“œ๋ฅผ ์ค„์ž…๋‹ˆ๋‹ค.
  2. ๊ณต๊ฐ„-์ฑ„๋„ ๋””์ปคํ”Œ๋ง ๋‹ค์šด ์ƒ˜ํ”Œ๋ง: ๊ณต๊ฐ„ ๊ฐ์†Œ์™€ ์ฑ„๋„ ๋ณ€์กฐ๋ฅผ ๋ถ„๋ฆฌํ•˜์—ฌ ์ •๋ณด ์†์‹ค๊ณผ ๊ณ„์‚ฐ ๋น„์šฉ์„ ์ตœ์†Œํ™”ํ•ฉ๋‹ˆ๋‹ค.
  3. ๋žญํฌ ๊ฐ€์ด๋“œ ๋ธ”๋ก ๋””์ž์ธ: ๋‚ด์žฌ์  ์Šคํ…Œ์ด์ง€ ๋ฆฌ๋˜๋˜์‹œ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๋ธ”๋ก ์„ค๊ณ„๋ฅผ ์กฐ์ •ํ•˜์—ฌ ์ตœ์ ์˜ ํŒŒ๋ผ๋ฏธํ„ฐ ํ™œ์šฉ์„ ๋ณด์žฅํ•ฉ๋‹ˆ๋‹ค.

์ •ํ™•๋„ ํ–ฅ์ƒ

  1. ๋Œ€ํ˜• ์ปค๋„ ์ปจ๋ณผ๋ฃจ์…˜: ์ˆ˜์šฉ ํ•„๋“œ๋ฅผ ํ™•๋Œ€ํ•˜์—ฌ ํŠน์ง• ์ถ”์ถœ ๊ธฐ๋Šฅ์„ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.
  2. ๋ถ€๋ถ„์  ์ž๊ธฐ ์ฃผ์˜(PSA): ์ตœ์†Œํ•œ์˜ ์˜ค๋ฒ„ํ—ค๋“œ๋กœ ๊ธ€๋กœ๋ฒŒ ํ‘œํ˜„ ํ•™์Šต์„ ๊ฐœ์„ ํ•˜๊ธฐ ์œ„ํ•ด ์ž๊ธฐ ์ฃผ์˜ ๋ชจ๋“ˆ์„ ํ†ตํ•ฉํ•ฉ๋‹ˆ๋‹ค.

์‹คํ—˜ ๋ฐ ๊ฒฐ๊ณผ

YOLOv10์€ COCO์™€ ๊ฐ™์€ ํ‘œ์ค€ ๋ฒค์น˜๋งˆํฌ์—์„œ ๊ด‘๋ฒ”์œ„ํ•˜๊ฒŒ ํ…Œ์ŠคํŠธ๋˜์–ด ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ๊ณผ ํšจ์œจ์„ฑ์„ ์ž…์ฆํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด ๋ชจ๋ธ์€ ๋‹ค์–‘ํ•œ ๋ณ€์ข…์— ๊ฑธ์ณ ์ตœ์ฒจ๋‹จ ๊ฒฐ๊ณผ๋ฅผ ๋‹ฌ์„ฑํ•˜์—ฌ ์ด์ „ ๋ฒ„์ „ ๋ฐ ๊ธฐํƒ€ ์ตœ์‹  ํƒ์ง€๊ธฐ์— ๋น„ํ•ด ์ง€์—ฐ ์‹œ๊ฐ„๊ณผ ์ •ํ™•๋„๊ฐ€ ํฌ๊ฒŒ ํ–ฅ์ƒ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.

๋น„๊ต

YOLOv10๊ณผ SOTA ๋ฌผ์ฒด ๊ฐ์ง€๊ธฐ ๋น„๊ต

๋‹ค๋ฅธ ์ตœ์ฒจ๋‹จ ํƒ์ง€๊ธฐ์™€ ๋น„๊ต:

  • YOLOv10-S / X๋Š” ๋น„์Šทํ•œ ์ •ํ™•๋„๋กœ RT-DETR-R18 / R101๋ณด๋‹ค 1.8๋ฐฐ / 1.3๋ฐฐ ๋น ๋ฆ…๋‹ˆ๋‹ค.
  • YOLOv10-B๋Š” ๋™์ผํ•œ ์ •ํ™•๋„์—์„œ YOLOv9-C๋ณด๋‹ค ํŒŒ๋ผ๋ฏธํ„ฐ ์ˆ˜๊ฐ€ 25% ๋” ์ ๊ณ  ์ง€์—ฐ ์‹œ๊ฐ„์ด 46% ๋” ์งง์Šต๋‹ˆ๋‹ค.
  • 1.8๋ฐฐ / 2.3๋ฐฐ ์ ์€ ๋งค๊ฐœ ๋ณ€์ˆ˜๋กœ YOLOv8-L / X๋ณด๋‹ค 0.3 AP / 0.5 AP ์„ฑ๋Šฅ์ด ์šฐ์ˆ˜ํ•ฉ๋‹ˆ๋‹ค.

๋‹ค์Œ์€ ๋‹ค๋ฅธ ์ตœ์‹  ๋ชจ๋ธ๊ณผ YOLOv10 ๋ณ€ํ˜•์„ ์ž์„ธํžˆ ๋น„๊ตํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค:

๋ชจ๋ธ ๋งค๊ฐœ๋ณ€์ˆ˜(M) ํ”Œ๋กญ(G) APval(%) ์ง€์—ฐ ์‹œ๊ฐ„(ms) ์ง€์—ฐ ์‹œ๊ฐ„(์ˆœ๋ฐฉํ–ฅ)(ms)
YOLOv6-3.0-N 4.7 11.4 37.0 2.69 1.76
๊ณจ๋“œ-YOLO-N 5.6 12.1 39.6 2.92 1.82
YOLOv8-N 3.2 8.7 37.3 6.16 1.77
YOLOv10-N 2.3 6.7 39.5 1.84 1.79
YOLOv6-3.0-S 18.5 45.3 44.3 3.42 2.35
Gold-YOLO-S 21.5 46.0 45.4 3.82 2.73
YOLOv8-S 11.2 28.6 44.9 7.07 2.33
YOLOv10-S 7.2 21.6 46.8 2.49 2.39
RT-DETR-R18 20.0 60.0 46.5 4.58 4.49
YOLOv6-3.0-M 34.9 85.8 49.1 5.63 4.56
Gold-YOLO-M 41.3 87.5 49.8 6.38 5.45
YOLOv8-M 25.9 78.9 50.6 9.50 5.09
YOLOv10-M 15.4 59.1 51.3 4.74 4.63
YOLOv6-3.0-L 59.6 150.7 51.8 9.02 7.90
๊ณจ๋“œ-YOLO-L 75.1 151.7 51.8 10.65 9.78
YOLOv8-L 43.7 165.2 52.9 12.39 8.06
RT-DETR-R50 42.0 136.0 53.1 9.20 9.07
YOLOv10-L 24.4 120.3 53.4 7.28 7.21
YOLOv8-X 68.2 257.8 53.9 16.86 12.83
RT-DETR-R101 76.0 259.0 54.3 13.71 13.58
YOLOv10-X 29.5 160.4 54.4 10.70 10.60

์‚ฌ์šฉ ์˜ˆ

YOLOv10์œผ๋กœ ์ƒˆ๋กœ์šด ์ด๋ฏธ์ง€๋ฅผ ์˜ˆ์ธกํ•ฉ๋‹ˆ๋‹ค:

from ultralytics import YOLO

# Load a pre-trained YOLOv10n model
model = YOLO("yolov10n.pt")

# Perform object detection on an image
results = model("image.jpg")

# Display the results
results[0].show()

์‚ฌ์šฉ์ž ์ง€์ • ๋ฐ์ดํ„ฐ ์„ธํŠธ์— ๋Œ€ํ•œ YOLOv10 ํ•™์Šต์šฉ:

from ultralytics import YOLO

# Load YOLOv10n model from scratch
model = YOLO("yolov10n.yaml")

# Train the model
model.train(data="coco8.yaml", epochs=100, imgsz=640)

์ง€์›๋˜๋Š” ์ž‘์—… ๋ฐ ๋ชจ๋“œ

The YOLOv10 models series offers a range of models, each optimized for high-performance Object Detection. These models cater to varying computational needs and accuracy requirements, making them versatile for a wide array of applications.

๋ชจ๋ธ ํŒŒ์ผ ์ด๋ฆ„ ์ž‘์—… ์ถ”๋ก  ์œ ํšจ์„ฑ ๊ฒ€์‚ฌ ๊ต์œก ๋‚ด๋ณด๋‚ด๊ธฐ
YOLOv10 yolov10n.pt yolov10s.pt yolov10m.pt yolov10l.pt yolov10x.pt ๋ฌผ์ฒด ๊ฐ์ง€ โœ… โœ… โœ… โœ…

Exporting YOLOv10

Due to the new operations introduced with YOLOv10, not all export formats provided by Ultralytics are currently supported. The following table outlines which formats have been successfully converted using Ultralytics for YOLOv10. Feel free to open a pull request if you're able to provide a contribution change for adding export support of additional formats for YOLOv10.

Export Format Supported
TorchScript โœ…
ONNX โœ…
OpenVINO โœ…
TensorRT โœ…
CoreML โŒ
TF SavedModel โŒ
TF GraphDef โŒ
TF Lite โŒ
TF Edge TPU โŒ
TF.js โŒ
PaddlePaddle โŒ
NCNN โŒ

๊ฒฐ๋ก 

YOLOv10์€ ์ด์ „ YOLO ๋ฒ„์ „์˜ ๋‹จ์ ์„ ํ•ด๊ฒฐํ•˜๊ณ  ํ˜์‹ ์ ์ธ ์„ค๊ณ„ ์ „๋žต์„ ํ†ตํ•ฉํ•˜์—ฌ ์‹ค์‹œ๊ฐ„ ๋ฌผ์ฒด ๊ฐ์ง€์˜ ์ƒˆ๋กœ์šด ํ‘œ์ค€์„ ์ œ์‹œํ•ฉ๋‹ˆ๋‹ค. ๋‚ฎ์€ ๊ณ„์‚ฐ ๋น„์šฉ์œผ๋กœ ๋†’์€ ์ •ํ™•๋„๋ฅผ ์ œ๊ณตํ•˜๋Š” ์ด ์ œํ’ˆ์€ ๋‹ค์–‘ํ•œ ์‹ค์ œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์— ์ด์ƒ์ ์ธ ์„ ํƒ์ž…๋‹ˆ๋‹ค.

์ธ์šฉ ๋ฐ ๊ฐ์‚ฌ

๊ด‘๋ฒ”์œ„ํ•œ ์—ฐ๊ตฌ์™€ ํ”„๋ ˆ์ž„์›Œํฌ์— ํฌ๊ฒŒ ๊ธฐ์—ฌํ•œ ์นญํ™”๋Œ€ํ•™๊ต์˜ YOLOv10 ์ €์ž๋“ค์—๊ฒŒ ๊ฐ์‚ฌ์˜ ๋ง์”€์„ ์ „ํ•ฉ๋‹ˆ๋‹ค. Ultralytics ํ”„๋ ˆ์ž„์›Œํฌ

@article{THU-MIGyolov10,
  title={YOLOv10: Real-Time End-to-End Object Detection},
  author={Ao Wang, Hui Chen, Lihao Liu, et al.},
  journal={arXiv preprint arXiv:2405.14458},
  year={2024},
  institution={Tsinghua University},
  license = {AGPL-3.0}
}

์ž์„ธํ•œ ๊ตฌํ˜„, ์•„ํ‚คํ…์ฒ˜ ํ˜์‹  ๋ฐ ์‹คํ—˜ ๊ฒฐ๊ณผ๋Š” ์นญํ™”๋Œ€ํ•™๊ต ํŒ€์˜ YOLOv10 ์—ฐ๊ตฌ ๋…ผ๋ฌธ๊ณผ GitHub ๋ฆฌํฌ์ง€ํ† ๋ฆฌ๋ฅผ ์ฐธ์กฐํ•˜์„ธ์š”.



Created 2024-05-25, Updated 2024-06-20
Authors: Burhan-Q (1), glenn-jocher (3), RizwanMunawar (2)

๋Œ“๊ธ€