çŸå£YOLOv6
æŠèŠ
Meituanã®YOLOv6ã¯ãã¹ããŒããšç²ŸåºŠã®ãã©ã³ã¹ã«åªããæå 端ã®ç©äœæ€åºåšã§ããããªã¢ã«ã¿ã€ã ã»ã¢ããªã±ãŒã·ã§ã³ã«åºãæ¡çšãããŠããŸãããã®ã¢ãã«ã§ã¯ãåæ¹åé£çµïŒBiCïŒã¢ãžã¥ãŒã«ã®å®è£ ãã¢ã³ã«ãŒæ¯æŽåŠç¿ïŒAATïŒæŠç¥ãCOCOããŒã¿ã»ããã§æå 端ã®ç²ŸåºŠãå®çŸããããã®æ¹è¯ãããããã¯ããŒã³ãšããã¯èšèšãªã©ãã¢ãŒããã¯ãã£ãšåŠç¿ã¹ããŒã ã«ããã€ãã®æ³šç®ãã¹ãæ¹è¯ãå ããããŠããŸãã
YOLOv6ã®æŠèŠã å€§å¹ ãªæ§èœåäžãããããããåèšèšããããããã¯ãŒã¯ã³ã³ããŒãã³ããšãã¬ãŒãã³ã°æŠç¥ã瀺ãã¢ãã«ã¢ãŒããã¯ãã£å³ã(aïŒYOLOv6ã®ããã¯ïŒNãšSã瀺ãïŒãM/Lã§ã¯RepBlocksãCSPStackRepã«çœ®ãæããããŠããã (b) BiCã¢ãžã¥ãŒã«ã®æ§é ã(c) SimCSPSPPFãããã¯ã(ãœãŒã¹).
äž»ãªç¹åŸŽ
- åæ¹åã³ã³ã«ãããŒã·ã§ã³ïŒBiCïŒã¢ãžã¥ãŒã«ïŒYOLOv6ã§ã¯ããã£ãã¯ã¿ã®ããã¯éšã«BiCã¢ãžã¥ãŒã«ãå°å ¥ããããŒã«ã©ã€ãŒãŒã·ã§ã³ä¿¡å·ã匷åããããšã§ãããããããªé床äœäžã§æ§èœãåäžãããŠããŸãã
- ã¢ã³ã«ãŒæ¯æŽãã¬ãŒãã³ã°ïŒAATïŒæŠç¥ïŒãã®ã¢ãã«ã¯ãæšè«å¹çãæãªãããšãªããã¢ã³ã«ãŒããŒã¹ãšã¢ã³ã«ãŒããªãŒã®äž¡æ¹ã®ãã©ãã€ã ã®å©ç¹ã享åããAATãææ¡ããã
- ããã¯ããŒã³ãšããã¯èšèšã®åŒ·åïŒYOLOv6ãæ·±åãããããã¯ããŒã³ãšããã¯ã«å¥ã®ã¹ããŒãžãå«ããããšã§ããã®ã¢ãã«ã¯é«è§£ååºŠå ¥åã®COCOããŒã¿ã»ããã§æå 端ã®æ§èœãéæããã
- èªå·±èžçæŠç¥ïŒYOLOv6ã®å°ããªã¢ãã«ã®æ§èœãåäžãããããã«ãæ°ããèªå·±èžçæŠç¥ãå®è£ ãããŠãããèšç·Žæã«è£å©çãªååž°åå²ã匷åããæšè«æã«ãããåé€ããããšã§ãèããé床äœäžãåé¿ããã
ããã©ãŒãã³ã¹ææš
YOLOv6ã«ã¯ãããŸããŸãªã¹ã±ãŒã«ã®äºååŠç¿æžã¿ã¢ãã«ãçšæãããŠããïŒ
- YOLOv6-N: COCO val2017ã§37.5%ã®APã1187FPSã§NVIDIA T4GPU ã
- YOLOv6-SïŒ484FPSã§45.0ïŒ ã®APã
- YOLOv6-MïŒ226FPSã§AP50.0ïŒ ã
- YOLOv6-LïŒ116FPSã§AP52.8ïŒ ã
- YOLOv6-L6ïŒæå 端ã®ç²ŸåºŠããªã¢ã«ã¿ã€ã ã§ã
YOLOv6ã¯ãŸããç°ãªã粟床ã®éååã¢ãã«ãã¢ãã€ã«ãã©ãããã©ãŒã ã«æé©åãããã¢ãã«ãæäŸããŠããã
䜿çšäŸ
ãã®äŸã§ã¯ãç°¡åãªYOLOv6ã®ãã¬ãŒãã³ã°ãšæšè«ã®äŸãæäŸããŸãããããã®ã¢ãŒããä»ã®ã¢ãŒãã«é¢ããå®å šãªããã¥ã¡ã³ãã¯ãPredict,Train,ValandExportdocsããŒãžãåç §ããŠãã ããã
äŸ
PyTorch ãã
ããããã *.pt
ã¢ãã«ããã³æ§æ *.yaml
ãã¡ã€ã«ã«æž¡ãããšãã§ããã YOLO()
ã¯ã©ã¹ã䜿çšããŠãpython ã«ã¢ãã«ã®ã€ã³ã¹ã¿ã³ã¹ãäœæããŸãïŒ
from ultralytics import YOLO
# Build a YOLOv6n model from scratch
model = YOLO("yolov6n.yaml")
# Display model information (optional)
model.info()
# Train the model on the COCO8 example dataset for 100 epochs
results = model.train(data="coco8.yaml", epochs=100, imgsz=640)
# Run inference with the YOLOv6n model on the 'bus.jpg' image
results = model("path/to/bus.jpg")
CLI ã³ãã³ãã§ã¢ãã«ãçŽæ¥å®è¡ã§ããïŒ
ãµããŒããããã¿ã¹ã¯ãšã¢ãŒã
YOLOv6ã·ãªãŒãºã«ã¯ãé«æ§èœãªç©äœæ€åºçšã«æé©åãããæ§ã ãªã¢ãã«ããããŸãããããã®ã¢ãã«ã¯ãããŸããŸãªèšç®ããŒãºã粟床èŠä»¶ã«å¯Ÿå¿ããå¹ åºãã¢ããªã±ãŒã·ã§ã³ã«å¯Ÿå¿ããŸãã
ã¢ãã«ã¿ã€ã | äºåã«èšç·ŽããããŠã§ã€ã | 察å¿ã¿ã¹ã¯ | æšè« | ããªããŒã·ã§ã³ | ãã¬ãŒãã³ã° | èŒžåº |
---|---|---|---|---|---|---|
YOLOv6-N | yolov6-n.pt |
ç©äœæ€åº | â | â | â | â |
YOLOv6-S | yolov6-s.pt |
ç©äœæ€åº | â | â | â | â |
YOLOv6-M | yolov6-m.pt |
ç©äœæ€åº | â | â | â | â |
YOLOv6-L | yolov6-l.pt |
ç©äœæ€åº | â | â | â | â |
YOLOv6-L6 | yolov6-l6.pt |
ç©äœæ€åº | â | â | â | â |
ãã®è¡šã¯ãYOLOv6ã¢ãã«ããªã¢ã³ãã®è©³çŽ°ãªæŠèŠã瀺ããŠãããç©äœæ€åºã¿ã¹ã¯ã«ãããããããã®æ©èœãšãæšè«ãæ€èšŒããã¬ãŒãã³ã°ããšã¯ã¹ããŒããšãã£ãæ§ã ãªæäœã¢ãŒããšã®äºææ§ã匷調ããŠããŸãããã®å æ¬çãªãµããŒãã«ããããŠãŒã¶ãŒã¯å¹ åºãç©äœæ€åºã·ããªãªã§YOLOv6ã¢ãã«ã®èœåããã«ã«æŽ»çšããããšãã§ããŸãã
åŒçšãšè¬èŸ
ãªã¢ã«ã¿ã€ã ç©äœæ€åºåéã«ãããèè ãã®å€å€§ãªè²¢ç®ã«è¬æãè¡šãããïŒ
ãªãªãžãã«ã®YOLOv6è«æã¯arXivã«æ²èŒãããŠãããèè ãã¯åœŒãã®ç 究ãå ¬éããã³ãŒãããŒã¹ã¯GitHubã§ã¢ã¯ã»ã¹ã§ãããæã ã¯ããã®åéãçºå±ãããããåºãã³ãã¥ããã£ã圌ãã®ç 究ã«ã¢ã¯ã»ã¹ã§ããããã«ãã圌ãã®åªåã«æè¬ããŠããã
ããããã質å
Meituanã®YOLOv6ãšã¯äœã§ããïŒ
Meituanã®YOLOv6ã¯ããªã¢ã«ã¿ã€ã ã¢ããªã±ãŒã·ã§ã³ã«çæ³çãªé床ãšç²ŸåºŠã®ãã©ã³ã¹ãåããæå 端ã®ãªããžã§ã¯ãæ€åºåšã§ããåæ¹åé£çµïŒBiCïŒã¢ãžã¥ãŒã«ãã¢ã³ã«ãŒæ¯æŽãã¬ãŒãã³ã°ïŒAATïŒã¹ãã©ããžãŒãªã©ã泚ç®ãã¹ãã¢ãŒããã¯ãã£ãŒã®åŒ·åãç¹åŸŽã§ãããããã®æè¡é©æ°ã«ãããé床å£åãæå°éã«æããªããå€§å¹ ãªæ§èœåäžãå®çŸããYOLOv6ã¯ç©äœæ€åºã¿ã¹ã¯ã«ãããŠç«¶äºåã®ããéžæè¢ãšãªã£ãŠããŸãã
YOLOv6ã®åæ¹åé£çµïŒBiCïŒã¢ãžã¥ãŒã«ã¯ã©ã®ããã«ããã©ãŒãã³ã¹ãåäžãããã®ãïŒ
YOLOv6ã®Bi-directional ConcatenationïŒBiCïŒã¢ãžã¥ãŒã«ã¯ãæ€åºåšã®ããã¯ã«ããããŒã«ãªãŒãŒã·ã§ã³ä¿¡å·ã匷åããé床ãžã®åœ±é¿ãæå°éã«æããªããããã©ãŒãã³ã¹ãåäžãããŸãããã®ã¢ãžã¥ãŒã«ã¯ãç°ãªãç¹åŸŽããããå¹æçã«çµåããç©äœãæ£ç¢ºã«æ€åºããã¢ãã«ã®èœåãåäžãããŸããYOLOv6ã®æ©èœã®è©³çŽ°ã«ã€ããŠã¯ãäž»èŠæ©èœã®ã»ã¯ã·ã§ã³ãåç §ããŠãã ããã
Ultralytics ã䜿ã£ãŠYOLOv6ã¢ãã«ããã¬ãŒãã³ã°ããã«ã¯ïŒ
YOLOv6ã¢ãã«ã¯ãUltralytics ã䜿ã£ãŠãPython ãŸãã¯CLI ã®ç°¡åãªã³ãã³ãã§ãã¬ãŒãã³ã°ã§ãããäŸãã°
äŸ
詳ããã¯é»è»ã®ããŒãžãã芧ãã ããã
YOLOv6ã®ããŸããŸãªããŒãžã§ã³ãšãã®ããã©ãŒãã³ã¹ææšã¯ïŒ
YOLOv6ã«ã¯è€æ°ã®ããŒãžã§ã³ããããããããç°ãªãããã©ãŒãã³ã¹èŠä»¶ã«æé©åãããŠããïŒ
- YOLOv6-NïŒ1187FPSã§37.5ïŒ ã®AP
- YOLOv6-SïŒ484FPSã§45.0ïŒ ã®AP
- YOLOv6-MïŒ226FPSã§AP50.0ïŒ ã
- YOLOv6-LïŒ116FPSã§52.8ïŒ ã®AP
- YOLOv6-L6ïŒãªã¢ã«ã¿ã€ã ã·ããªãªã«ãããæå 端ã®ç²ŸåºŠ
ãããã®ã¢ãã«ã¯ãNVIDIA T4GPU ãCOCOããŒã¿ã»ããã§è©äŸ¡ãããŠãããããã©ãŒãã³ã¹ã»ã¡ããªã¯ã¹ã®è©³çŽ°ã«ã€ããŠã¯ãããã©ãŒãã³ã¹ã»ã¡ããªã¯ã¹ã®ã»ã¯ã·ã§ã³ãåç §ã
ã¢ã³ã«ãŒæ¯æŽãã¬ãŒãã³ã°ïŒAATïŒæŠç¥ã¯YOLOv6ã«ã©ã®ãããªã¡ãªãããããããã®ãïŒ
YOLOv6ã®ã¢ã³ã«ãŒæ¯æŽãã¬ãŒãã³ã°ïŒAATïŒã¯ãã¢ã³ã«ãŒããŒã¹ãšã¢ã³ã«ãŒããªãŒã®ã¢ãããŒãã®èŠçŽ ãçµã¿åãããæšè«å¹çãæãªãããšãªãã¢ãã«ã®æ€åºèœåãåäžãããããã®æŠç¥ã¯ãåŠç¿äžã«ã¢ã³ã«ãŒã掻çšããŠããŠã³ãã£ã³ã°ããã¯ã¹ã®äºæž¬ãåäžãããYOLOv6ãå€æ§ãªç©äœæ€åºã¿ã¹ã¯ã«æå¹ãªãã®ã«ããŸãã
Ultralytics ãYOLOv6ã¢ãã«ããµããŒãããŠããéçšã¢ãŒãã¯ïŒ
YOLOv6ã¯ãæšè«ãæ€èšŒããã¬ãŒãã³ã°ããšã¯ã¹ããŒããå«ãæ§ã ãªæäœã¢ãŒãããµããŒãããŠããããã®æè»æ§ã«ããããŠãŒã¶ãŒã¯æ§ã ãªã·ããªãªã§ã¢ãã«ã®èœåããã«ã«æŽ»çšããããšãã§ãããåã¢ãŒãã®è©³çŽ°ã«ã€ããŠã¯ãããµããŒããããã¿ã¹ã¯ãšã¢ãŒããã®ã»ã¯ã·ã§ã³ãã芧ãã ããã