YOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ã«Kaggleã䜿çšããã¬ã€ã
AIã«ã€ããŠåŠã³ãå°èŠæš¡ãªãããžã§ã¯ãã«åãçµãã§ããå Žåã匷åãªã³ã³ãã¥ãŒãã£ã³ã°ã»ãªãœãŒã¹ããŸã å©çšã§ããªããããããªããããã€ãšã³ãã®ããŒããŠã§ã¢ã¯ããªãé«äŸ¡ã§ããã幞ããªããšã«ãGoogle ãææãããã©ãããã©ãŒã ãKaggle ã¯çŽ æŽããããœãªã¥ãŒã·ã§ã³ãæäŸããŠãããKaggleã¯ç¡æã®ã¯ã©ãŠãããŒã¹ã®ç°å¢ãæäŸããŠãããGPU ã®ãªãœãŒã¹ã«ã¢ã¯ã»ã¹ãããã倧èŠæš¡ãªããŒã¿ã»ãããæ±ã£ãããããŒã¿ãµã€ãšã³ãã£ã¹ããæ©æ¢°åŠç¿æ奜家ã®å€æ§ãªã³ãã¥ããã£ãšã³ã©ãã¬ãŒã·ã§ã³ãããããããšãã§ããã
Kaggleã¯ãã¢ãã«ã®ãã¬ãŒãã³ã°ãå®éšã«æé©ã§ãã Ultralytics YOLOv8ã¢ãã«ã®ãã¬ãŒãã³ã°ãå®éšã«æé©ã§ããKaggleããŒãããã¯ã䜿ãã°ã人æ°ã®æ©æ¢°åŠç¿ã©ã€ãã©ãªããã¬ãŒã ã¯ãŒã¯ããããžã§ã¯ãã§ç°¡åã«äœ¿ãããšãã§ããŸããKaggleã®äž»ãªæ©èœãæ¢çŽ¢ãããã®ãã©ãããã©ãŒã ã§YOLOv8 ã¢ãã«ããã¬ãŒãã³ã°ããæ¹æ³ãåŠã³ãŸãããïŒ
Kaggleãšã¯ïŒ
Kaggleã¯ãäžçäžã®ããŒã¿ãµã€ãšã³ãã£ã¹ããéãŸããå®éã®ããŒã¿ãµã€ãšã³ã¹ã®åé¡ã解決ããããã«ååããåŠã³ã競ãåããã©ãããã©ãŒã ã§ããAnthony GoldbloomãšJeremy Howardã«ãã£ãŠ2010幎ã«ç«ã¡äžãããã2017幎ã«Google ãKaggleã¯ããŠãŒã¶ãŒãããŒã¿ã»ãããæ¥ç¶ãçºèŠãå ±æããGPU-powered notebooksã䜿çšããããŒã¿ãµã€ãšã³ã¹ç«¶æäŒã«åå ããããšãå¯èœã«ããããã®ãã©ãããã©ãŒã ã¯ãå ç¢ãªããŒã«ãšãªãœãŒã¹ãæäŸããããšã§ãçç·Žããå°é家ãšç±å¿ãªåŠç¿è ã®äž¡æ¹ãç®æšãéæã§ããããã«èšèšãããŠããã
2022幎çŸåš1,000äžäººä»¥äžã®ãŠãŒã¶ãŒãæã€Kaggleã¯ãæ©æ¢°åŠç¿ã¢ãã«ã®éçºãšå®éšã®ããã®è±å¯ãªç°å¢ãæäŸããŠããŸããããŒã«ã«ãã·ã³ã®ã¹ããã¯ãã»ããã¢ãããå¿é ããå¿ èŠã¯ãªããKaggleã®ã¢ã«ãŠã³ããšãŠã§ããã©ãŠã¶ãããã°ããã«å©çšããããšãã§ããŸãã
Kaggle ã䜿ã£ããã¬ãŒãã³ã°YOLOv8
Kaggleã§ã®YOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ã¯ããã©ãããã©ãŒã ã匷åãªGPUãå©çšã§ãããããã·ã³ãã«ã§å¹ççã§ãã
å§ããã«ã¯ãKaggleYOLOv8 Notebookã«ã¢ã¯ã»ã¹ãããKaggleã®ç°å¢ã«ã¯ãTensorFlow ãPyTorch ã®ãããªã©ã€ãã©ãªãããªã€ã³ã¹ããŒã«ãããŠãããã»ããã¢ããã«æéãããããŸããã
Kaggleã¢ã«ãŠã³ãã«ãµã€ã³ã€ã³ãããšãã³ãŒããã³ããŒããŠç·šéãããªãã·ã§ã³ãã¯ãªãã¯ããã¢ã¯ã»ã©ã¬ãŒã¿èšå®ã§GPU ãããŒãããã¯ã®ã»ã«ãå®è¡ããŠã¢ãã«ã®ãã¬ãŒãã³ã°ãéå§ããããšãã§ããŸããã¢ãã«ãã¬ãŒãã³ã°ããã»ã¹ãšãã¹ããã©ã¯ãã£ã¹ã®è©³çŽ°ã«ã€ããŠã¯ãYOLOv8 ã¢ãã«ãã¬ãŒãã³ã°ã¬ã€ããåç §ããŠãã ããã
YOLOv8 Kaggleã®å ¬åŒããŒãããã¯ããŒãžã§ãå³äžé ã«ãã3ã€ã®ç¹ãã¯ãªãã¯ãããšãããå€ãã®ãªãã·ã§ã³ããããã¢ãã衚瀺ãããŸãã
ãããã®ãªãã·ã§ã³ã«ã¯ä»¥äžãå«ãŸããïŒ
- ããŒãžã§ã³ãèŠãïŒããŒãããã¯ã®æ§ã ãªããŒãžã§ã³ããã©ãŠãºããŠãæéã®çµéã«äŒŽãå€æŽã確èªããå¿ èŠã«å¿ããŠä»¥åã®ããŒãžã§ã³ã«æ»ãããšãã§ããŸãã
- APIã³ãã³ãã®ã³ããŒïŒèªååãã¯ãŒã¯ãããŒãžã®çµ±åã«äŸ¿å©ã§ãã
- Google ããŒãããã¯ã§éãïŒGoogle ã®ãã¹ããããããŒãããã¯ç°å¢ã§ããŒãããã¯ãéããŸãã
- Colabã§éãïŒGoogle Colabã§ããŒãããã¯ãèµ·åããç·šéãå®è¡ãã§ããŸãã
- ã³ã¡ã³ãããã©ããŒããã³ã¡ã³ãæ¬ã賌èªããŠãææ°æ å ±ãå ¥æããã³ãã¥ããã£ãšäº€æµããŸãããã
- ããŠã³ããŒãã³ãŒãããŒãããã¯å šäœãJupyter (.ipynb)ãã¡ã€ã«ãšããŠããŠã³ããŒããããªãã©ã€ã³ã§äœ¿çšããããããŒã«ã«ç°å¢ã§ããŒãžã§ã³ç®¡çãããããããšãã§ããŸãã
- ã³ã¬ã¯ã·ã§ã³ã«è¿œå Kaggleã¢ã«ãŠã³ãå ã®ã³ã¬ã¯ã·ã§ã³ã«ããŒãããã¯ãä¿åããç°¡åã«ã¢ã¯ã»ã¹ã»æŽçããããšãã§ããŸãã
- ããã¯ããŒã¯ããŒãããã¯ãããã¯ããŒã¯ããŠãå°æ¥ããã«ã¢ã¯ã»ã¹ã§ããããã«ããŸãã
- ããŒãããã¯ãåã蟌ãïŒããã°ããŠã§ããµã€ããããã¥ã¡ã³ãã«ããŒãããã¯ãå«ããããã®åã蟌ã¿ãªã³ã¯ãååŸããŸãã
Kaggleã§ã®äœæ¥äžã«ããããåé¡
Kaggleã§ä»äºãããŠãããšãããããåé¡ã«åºãããããšããããŸããããã§ã¯ãKaggleãã¹ã ãŒãºã«å©çšããããã®ãã€ã³ããã玹ä»ããŸãïŒ
- GPUãžã®ã¢ã¯ã»ã¹ïŒKaggleããŒãããã¯ã§ã¯ããã€ã§ãGPU ãé±ã«30æéãŸã§äœ¿çšããããšãã§ããŸããKaggleã¯16GBã®ã¡ã¢ãªãæèŒããNvidia Tesla P100GPU ãæäŸããNvidia GPU T4 x2ã䜿çšãããªãã·ã§ã³ãæäŸããŠããŸãã匷åãªããŒããŠã§ã¢ã¯ãæ©æ¢°åŠç¿ã¿ã¹ã¯ãå éããã¢ãã«ã®ãã¬ãŒãã³ã°ãšæšè«ãã¯ããã«é«éåããŸãã
- Kaggle KernelsïŒKaggle Kernelsã¯GPUãçµ±åã§ããç¡æã®JupyterããŒãããã¯ãµãŒããŒã§ãã¯ã©ãŠãã³ã³ãã¥ãŒã¿ãŒäžã§æ©æ¢°åŠç¿æäœãå®è¡ããããšãã§ãããèªåã®ã³ã³ãã¥ãŒã¿ã®CPU ãéè² è·ãåé¿ããããŒã«ã«ãªãœãŒã¹ã解æŸããããšãã§ããŸãã
- KaggleããŒã¿ã»ããKaggleã®ããŒã¿ã»ããã¯ç¡æã§ããŠã³ããŒãã§ããŸãããã ããåããŒã¿ã»ããã®ã©ã€ã»ã³ã¹ã確èªãã䜿çšäžã®å¶éãç解ããããšãéèŠã§ããããŒã¿ã»ããã«ãã£ãŠã¯ãåŠè¡è«æãåçšå©çšãå¶éãããŠããå ŽåããããŸããããŒã¿ã»ããã¯KaggleããŒãããã¯ã«çŽæ¥ããŠã³ããŒãããããšãã§ããŸãããKaggle APIãéããŠä»ã®å Žæã«ããŠã³ããŒãããããšãã§ããŸãã
- ããŒãããã¯ã®ä¿åãšã³ãããKaggleã§ããŒãããã¯ãä¿åãã³ãããããã«ã¯ã"Save Version "ãã¯ãªãã¯ããŸããããã§ããŒãããã¯ã®çŸåšã®ç¶æ ãä¿åãããŸããããã¯ã°ã©ãŠã³ãã«ãŒãã«ãåºåãã¡ã€ã«ãçæãçµãããšãã¡ã€ã³ããŒãããã¯ããŒãžã®åºåã¿ãããåºåãã¡ã€ã«ã«ã¢ã¯ã»ã¹ããããšãã§ããŸãã
- ã³ã©ãã¬ãŒã·ã§ã³Kaggleã¯ã³ã©ãã¬ãŒã·ã§ã³ããµããŒãããŠããŸãããè€æ°ã®ãŠãŒã¶ãŒãåæã«ããŒãããã¯ãç·šéããããšã¯ã§ããŸãããKaggleã§ã®ã³ã©ãã¬ãŒã·ã§ã³ã¯éåæã§è¡ãããŸããã€ãŸãããŠãŒã¶ãŒã¯åãããŒãããã¯ãç°ãªãæéã«å ±æããäœæ¥ããããšãã§ããŸãã
- 以åã®ããŒãžã§ã³ã«æ»ãããŒãããã¯ã®ä»¥åã®ããŒãžã§ã³ã«æ»ãå¿ èŠãããå ŽåãããŒãããã¯ãéããå³äžã®çžŠã«äžŠãã 3ã€ã®ç¹ãã¯ãªãã¯ããŠã"View Versions "ãéžæããŸããæ»ãããããŒãžã§ã³ãèŠã€ãããã®æšªã«ãã"... "ã¡ãã¥ãŒãã¯ãªãã¯ãã"ããŒãžã§ã³ã«æ»ã "ãéžæããŸããããŒãããã¯ãå ã«æ»ã£ããã"Save Version "ãã¯ãªãã¯ããŠå€æŽãã³ãããããŸãã
Kaggleã®äž»ãªç¹åŸŽ
次ã«ãKaggleãããŒã¿ãµã€ãšã³ã¹ãæ©æ¢°åŠç¿ã®æ奜家ã«ãšã£ãŠåªãããã©ãããã©ãŒã ã§ããçç±ãç解ããŸããããäž»ãªãã€ã©ã€ããããã€ã玹ä»ãããïŒ
- ããŒã¿ã»ããKaggleã¯æ§ã ãªãããã¯ã«é¢ããèšå€§ãªããŒã¿ã»ããããã¹ãããŠããŸãããããã®ããŒã¿ã»ãããç°¡åã«æ€çŽ¢ãããããžã§ã¯ãã§äœ¿çšããããšãã§ããŸããYOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ããã¹ãã«ç¹ã«äŸ¿å©ã§ãã
- ã³ã³ããã£ã·ã§ã³ãšããµã€ãã£ã³ã°ãªã³ã³ããã£ã·ã§ã³ã§ç¥ãããKaggleã§ã¯ãããŒã¿ãµã€ãšã³ãã£ã¹ããæ©æ¢°åŠç¿æ奜家ãå®äžçã®åé¡ã解決ããããšãã§ããŸããã³ã³ããã£ã·ã§ã³ã«åå ããããšã§ãã¹ãã«ãåäžãããæ°ãããã¯ããã¯ãåŠã³ãã³ãã¥ããã£ã§èªããããããã«ãªããŸãã
- TPUãžã®ç¡æã¢ã¯ã»ã¹Kaggleã¯ãè€éãªæ©æ¢°åŠç¿ã¢ãã«ã®ãã¬ãŒãã³ã°ã«äžå¯æ¬ ãªåŒ·åãªTPUãžã®ç¡æã¢ã¯ã»ã¹ãæäŸããŠããŸããã€ãŸããäœèšãªã³ã¹ãããããããšãªããYOLOv8 ãããžã§ã¯ãã®åŠçãã¹ããŒãã¢ããããããã©ãŒãã³ã¹ãé«ããããšãã§ããŸãã
- Githubãšã®çµ±åïŒKaggleã§ã¯ãããŒãããã¯ãã¢ããããŒãããäœæ¥ãä¿åããããã«GitHubãªããžããªã«ç°¡åã«æ¥ç¶ããããšãã§ããŸãããã®çµ±åã«ããããã¡ã€ã«ã®ç®¡çãšã¢ã¯ã»ã¹ã䟿å©ã«ãªããŸãã
- ã³ãã¥ããã£ãšãã£ã¹ã«ãã·ã§ã³Kaggleã¯ãããŒã¿ãµã€ãšã³ãã£ã¹ããšæ©æ¢°åŠç¿å®è·µè ã®åŒ·åãªã³ãã¥ããã£ãèªã£ãŠããŸãããã£ã¹ã«ãã·ã§ã³ãã©ãŒã©ã ãšå ±æããŒãããã¯ã¯ãåŠç¿ãšãã©ãã«ã·ã¥ãŒãã£ã³ã°ã®ããã®çŽ æŽããããªãœãŒã¹ã§ããç°¡åã«ãã«ããèŠã€ããç¥èãå ±æããä»ã®äººãšååããããšãã§ããŸãã
YOLOv8 ããªãKaggleã䜿ãã¹ããªã®ãïŒ
æ©æ¢°åŠç¿ã¢ãã«ããã¬ãŒãã³ã°ã»è©äŸ¡ããããã®ãã©ãããã©ãŒã ã¯è€æ°ãããŸãããKaggleã®ç¹åŸŽã¯äœã§ããããïŒããã§ã¯ãæ©æ¢°åŠç¿ãããžã§ã¯ãã«Kaggleã䜿çšããã¡ãªãããã玹ä»ããŸãããïŒ
- å ¬éããŒãããã¯ïŒKaggleããŒãããã¯ãå ¬éããããšã§ãä»ã®ãŠãŒã¶ãŒãããªãã®ããŒãããã¯ãé²èŠ§ãæ祚ããã©ãŒã¯ããã£ã¹ã«ãã·ã§ã³ããããšãã§ããŸããKaggleã¯ã³ã©ãã¬ãŒã·ã§ã³ããã£ãŒãããã¯ãã¢ã€ãã¢ã®å ±æãä¿é²ããããªãã®YOLOv8 ã¢ãã«ã®æ¹åã«åœ¹ç«ã¡ãŸãã
- ããŒãããã¯ã³ãããã®å æ¬çãªå±¥æŽïŒKaggleã¯ããªãã®ããŒãããã¯ã®ã³ãããã®è©³çŽ°ãªå±¥æŽãäœæããŸããããã«ãããæéã®çµéã«äŒŽãå€æŽã確èªã»è¿œè·¡ããããšãã§ãããããžã§ã¯ãã®é²åãç解ããå¿ èŠã«å¿ããŠä»¥åã®ããŒãžã§ã³ã«æ»ãããšã容æã«ãªããŸãã
- ã³ã³ãœãŒã«ãžã®ã¢ã¯ã»ã¹Kaggleã¯ã³ã³ãœãŒã«ãæäŸããããªãã®ç°å¢ãããã³ã³ãããŒã«ã§ããããã«ããŸãããã®æ©èœã«ãããã³ãã³ãã©ã€ã³ããçŽæ¥æ§ã ãªã¿ã¹ã¯ãå®è¡ããããšãã§ããã¯ãŒã¯ãããŒãšçç£æ§ãåäžããŸãã
- ãªãœãŒã¹ã®å©çšKaggleã®åããŒãããã¯ç·šéã»ãã·ã§ã³ã«ã¯ãããªãã®ãªãœãŒã¹ãæäŸãããŸãïŒCPU ãGPU ã»ãã·ã§ã³ã®å®è¡æéã¯12æéãTPU ã»ãã·ã§ã³ã®å®è¡æéã¯9æéãèªåä¿åããããã£ã¹ã¯å®¹éã¯20ã®ã¬ãã€ãã§ãã
- ããŒãããã¯ã®ã¹ã±ãžã¥ãŒãªã³ã°Kaggleã§ã¯ãç¹å®ã®æéã«ããŒãããã¯ãå®è¡ããããã¹ã±ãžã¥ãŒãªã³ã°ããããšãã§ããŸããå®æçãªã¢ãã«ã®ãã¬ãŒãã³ã°ã®ãããªç¹°ãè¿ãã®ã¿ã¹ã¯ãæåã§è¡ãããšãªãèªååããããšãã§ããŸãã
Kaggleã«ã€ããŠåŠã³ç¶ãã
Kaggleã«ã€ããŠãã詳ããç¥ãããæ¹ã¯ããã¡ããåèã«ããŠãã ããïŒ
- ã«ã°ã«ã»ã©ãŒã³:Kaggle Learnã§æ§ã ãªç¡æã®ã€ã³ã¿ã©ã¯ãã£ããã¥ãŒããªã¢ã«ãã芧ãã ããããããã®ã³ãŒã¹ã§ã¯ãããŒã¿ãµã€ãšã³ã¹ã«äžå¯æ¬ ãªãããã¯ãåãäžããæ°ããã¹ãã«ã®ç¿åŸã«åœ¹ç«ã€å®è·µçãªäœéšãæäŸããŸãã
- Kaggleå ¥é:ãã®å æ¬çãªã¬ã€ãã¯ãã³ã³ããã£ã·ã§ã³ãžã®åå ããæåã®ããŒãããã¯ã®äœæãŸã§ãKaggleã®äœ¿ãæ¹ã®åºæ¬ã説æããŸããåå¿è ã®æ¹ã«ãšã£ãŠçŽ æŽãããåºçºç¹ã§ãã
- Kaggle Medium ããŒãž:Kaggleã®MediumããŒãžã§ã¯ããã¥ãŒããªã¢ã«ãã¢ããããŒããã³ãã¥ããã£ãžã®æçš¿ãã芧ããã ããŸããææ°ã®ãã¬ã³ããææ¡ããããŒã¿ãµã€ãšã³ã¹ãžã®æ·±ãæŽå¯ãåŸãããã®åªããæ å ±æºã§ãã
æŠèŠ
æã ã¯ãKaggleã匷åãªGPUãžã®ç¡æã¢ã¯ã»ã¹ãæäŸããã¢ãã«ã®ãã¬ãŒãã³ã°ãšè©äŸ¡ãå¹çåããããšã§ãYOLOv8 ãããžã§ã¯ããåŸæŒãã§ããããšãèŠãŠããŸãããKaggleã®ãã©ãããã©ãŒã ã¯ãŠãŒã¶ãŒãã¬ã³ããªãŒã§ãè¿ éãªã»ããã¢ããã®ããã®ã©ã€ãã©ãªãããªã€ã³ã¹ããŒã«ãããŠããŸãã
詳现ã¯Kaggleã®ããã¥ã¡ã³ããã芧ãã ããã
YOLOv8 ã®ãã®ä»ã®çµ±åã«èå³ããããŸããïŒ Ultralytics çµ±åã¬ã€ãããã§ãã¯ããŠãæ©æ¢°åŠç¿ãããžã§ã¯ãã®ããã®è¿œå ã®ããŒã«ãæ©èœã調ã¹ãŠãã ããã
ããããã質å
Kaggleã§YOLOv8 ã¢ãã«ããã¬ãŒãã³ã°ããã«ã¯ïŒ
Kaggleã§ã®YOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ã¯ç°¡åã§ããããŸããKaggleYOLOv8 ããŒãããã¯ã«ã¢ã¯ã»ã¹ããŸããKaggleã¢ã«ãŠã³ãã«ãµã€ã³ã€ã³ããããŒãããã¯ãã³ããŒããŠç·šéããã¢ã¯ã»ã©ã¬ãŒã¿èšå®ã§GPU ãããŒãããã¯ã®ã»ã«ãå®è¡ããŠãã¬ãŒãã³ã°ãéå§ããŸãããã詳现ãªæé ã«ã€ããŠã¯ãYOLOv8 ã¢ãã«ãã¬ãŒãã³ã°ã¬ã€ããåç §ããŠãã ããã
YOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ã«Kaggleã䜿ãã¡ãªããã¯äœã§ããïŒ
Kaggleã¯ãYOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ã«ããã€ãã®å©ç¹ãæäŸããŠããïŒ
- GPU ç¡æã¢ã¯ã»ã¹ïŒNvidia Tesla P100ãT4 x2ãªã©ã®åŒ·åãªGPUãé±30æéãŸã§å©çšã§ããŸãã
- ããªã€ã³ã¹ããŒã«ã»ã©ã€ãã©ãªïŒTensorFlow ãPyTorch ãªã©ã®ã©ã€ãã©ãªãããªã€ã³ã¹ããŒã«ãããŠãããã»ããã¢ãããç°¡åã§ãã
- ã³ãã¥ããã£ãšã®ã³ã©ãã¬ãŒã·ã§ã³ïŒããŒã¿ãµã€ãšã³ãã£ã¹ããæ©æ¢°åŠç¿æ奜家ã®åºå€§ãªã³ãã¥ããã£ã«åå ã§ããŸãã
- ããŒãžã§ã³ç®¡çïŒããŒãããã¯ã®ç°ãªãããŒãžã§ã³ãç°¡åã«ç®¡çããå¿ èŠã«å¿ããŠä»¥åã®ããŒãžã§ã³ã«æ»ãããšãã§ããŸãã
詳ããã¯ãUltralytics çµ±åã¬ã€ããã芧ãã ããã
Kaggle forYOLOv8 ã䜿çšããéãã©ã®ãããªäžè¬çãªåé¡ã«ééããå¯èœæ§ããããã©ã®ããã«è§£æ±ºããã°ããã§ããããïŒ
ããããåé¡ã¯ä»¥äžã®éãïŒ
- GPUãžã®ã¢ã¯ã»ã¹ïŒããŒãããã¯ã®èšå®ã§GPU ãæå¹ã«ããŠãã ãããKaggleã§ã¯ãGPU ãé±ã«30æéãŸã§äœ¿çšã§ããŸãã
- ããŒã¿ã»ããã®ã©ã€ã»ã³ã¹åããŒã¿ã»ããã®ã©ã€ã»ã³ã¹ã確èªãã䜿çšå¶éãç解ããŠãã ããã
- ããŒãããã¯ã®ä¿åãšã³ãããSave Version "ãã¯ãªãã¯ãããšãããŒãããã¯ã®ç¶æ ãä¿åãããOutputã¿ãããåºåãã¡ã€ã«ã«ã¢ã¯ã»ã¹ã§ããŸãã
- ã³ã©ãã¬ãŒã·ã§ã³Kaggleã¯éåæã®ã³ã©ãã¬ãŒã·ã§ã³ããµããŒãããŠããŸããè€æ°ã®ãŠãŒã¶ãŒãåæã«ããŒãããã¯ãç·šéããããšã¯ã§ããŸããã
ãã©ãã«ã·ã¥ãŒãã£ã³ã°ã®ãã³ãã«ã€ããŠã¯ããããããåé¡ãã¬ã€ããã芧ãã ããã
YOLOv8 ã¢ãã«ã®ãã¬ãŒãã³ã°ã«ãGoogle Colab ã®ãããªä»ã®ãã©ãããã©ãŒã ã§ã¯ãªããKaggle ãéžã¶ã¹ãçç±ã¯äœã§ããããïŒ
Kaggleã¯ãåªããéžæè¢ãšãªããŠããŒã¯ãªæ©èœãæäŸããŠããŸãïŒ
- å ¬éããŒãããã¯ïŒããªãã®äœåãã³ãã¥ããã£ã§å ±æãããã£ãŒãããã¯ãã³ã©ãã¬ãŒã·ã§ã³ãããŸãããã
- TPUãžã®ç¡æã¢ã¯ã»ã¹ïŒè¿œå è²»çšãªãã§ããã¯ãã«ãªTPUã§ãã¬ãŒãã³ã°ãã¹ããŒãã¢ããã
- å æ¬çãªå±¥æŽïŒããŒãããã¯ã®ã³ãããã®è©³çŽ°ãªå±¥æŽã§ãé·æã«ãããå€æŽã远跡ã§ããŸãã
- ãªãœãŒã¹ã®å©çšå¯èœæ§ïŒåããŒãããã¯ã»ãã·ã§ã³ã«ã¯ãCPU ãšGPU ã»ãã·ã§ã³ã®12æéã®å®è¡æéãå«ããããªãã®ãªãœãŒã¹ãæäŸãããŸãã Google Colabãšã®æ¯èŒã«ã€ããŠã¯ãGoogle Colabã¬ã€ãããåç §ãã ããã
KaggleããŒãããã¯ã以åã®ããŒãžã§ã³ã«æ»ãã«ã¯ã©ãããã°ããã§ããïŒ
以åã®ããŒãžã§ã³ã«æ»ãã«ã¯
- ããŒãããã¯ãéããå³äžã®çžŠã«äžŠãã 3ã€ã®ç¹ãã¯ãªãã¯ããã
- "ããŒãžã§ã³ãèŠã "ãéžæããã
- æ»ãããããŒãžã§ã³ãèŠã€ãããã®æšªã«ãã"... "ã¡ãã¥ãŒãã¯ãªãã¯ãã"ããŒãžã§ã³ãæ»ã "ãéžæããã
- Save Version "ãã¯ãªãã¯ããŠå€æŽãã³ãããããã