ã€ã³ãã«OpenVINO ãšã¯ã¹ããŒã
ãã®ã¬ã€ãã§ã¯ãYOLOv8 ã¢ãã«ã OpenVINOãã©ãŒããããžã®ã¢ãã«ã®ãšã¯ã¹ããŒãã«ã€ããŠèª¬æããããã®ãã©ãŒãããã§ã¯ãCPUãæ倧3åé«éåãããã ãã§ãªããIntelGPUã NPUããŒããŠã§ã¢äžã§YOLO æšè«ãé«éåãããã
OpenVINOOpen Visual Inference & Neural Network Optimization toolkitã®ç¥ã§ãAIæšè«ã¢ãã«ã®æé©åãšå±éã®ããã®å æ¬çãªããŒã«ãããã§ãããååã«Visualãå«ãŸããŠããŸãããOpenVINO ãèšèªãé³å£°ãæç³»åãªã©æ§ã ãªè¿œå ã¿ã¹ã¯ããµããŒãããŠããŸãã
èŠããã ïŒ OpenVINO ã䜿ã£ãæšè«ã®ããã«Ultralytics YOLOv8 ã¢ãã«ããšã¯ã¹ããŒãããŠæé©åããæ¹æ³ .
䜿çšäŸ
YOLOv8n ã¢ãã«ãOpenVINO 圢åŒã«ãšã¯ã¹ããŒããããšã¯ã¹ããŒãããã¢ãã«ã§æšè«ãå®è¡ããã
äŸ
from ultralytics import YOLO
# Load a YOLOv8n PyTorch model
model = YOLO('yolov8n.pt')
# Export the model
model.export(format='openvino') # creates 'yolov8n_openvino_model/'
# Load the exported OpenVINO model
ov_model = YOLO('yolov8n_openvino_model/')
# Run inference
results = ov_model('https://ultralytics.com/images/bus.jpg')
è«äº
ã㌠| äŸ¡å€ | 説æ |
---|---|---|
format |
'openvino' |
ãšã¯ã¹ããŒããããã©ãŒããã |
imgsz |
640 |
ã¹ã«ã©ãŒãŸã㯠(h, w) ãªã¹ããšããŠã®ç»åãµã€ãºïŒããªãã¡ (640, 480) |
half |
False |
FP16éåå |
ã¡ãªããOpenVINO
- ããã©ãŒãã³ã¹ïŒOpenVINO ã¯ãã€ã³ãã«CPUãçµ±åGPUããã£ã¹ã¯ãªãŒãGPUãFPGAã®ãã¯ãŒã掻çšããããšã§ãé«æ§èœãªæšè«ãå®çŸããŸãã
- ããããžãã¢ã¹å®è¡ã®ãµããŒã:OpenVINO ã¯ãäžåºŠæžãã°ããµããŒããããŠããããããã€ã³ãã«ã»ããŒããŠã§ã¢ïŒCPUãGPUãFPGAãVPUãªã©ïŒã«ãããã€ã§ããAPIãæäŸããŸãã
- Model Optimizer:OpenVINO ã¯ãPyTorch ãTensorFlow ãTensorFlow LiteãKerasãONNX ãPaddlePaddle ãCaffe ãªã©ã®äžè¬çãªãã£ãŒãã©ãŒãã³ã°ãã¬ãŒã ã¯ãŒã¯ããã¢ãã«ãã€ã³ããŒããå€æãæé©åãã Model Optimizer ãæäŸããŸãã
- 䜿ããããïŒããŒã«ãããã«ã¯ãããŒã«ãããã®ããŸããŸãªåŽé¢ãæãã80以äžã®ãã¥ãŒããªã¢ã«ããŒããã㯠ïŒYOLOv8 æé©åãå«ãïŒãä»å±ããŠããŸãã
OpenVINO 茞åºæ§é
ã¢ãã«ãOpenVINO ãã©ãŒãããã«ãšã¯ã¹ããŒããããšã次ã®ãããªãã£ã¬ã¯ããªãäœæãããŸãïŒ
- XMLãã¡ã€ã«ïŒãããã¯ãŒã¯ã®ããããžãŒãèšè¿°ããã
- BINãã¡ã€ã«ïŒweights and biases ãã€ããªããŒã¿ãå«ãã
- ãããã³ã°ãã¡ã€ã«ïŒå ã®ã¢ãã«åºåãã³ãœã«ã®OpenVINO tensor ãžã®ãããã³ã°ãä¿æããã
ãããã®ãã¡ã€ã«ã䜿çšããŠãOpenVINO æšè«ãšã³ãžã³ã§æšè«ãå®è¡ããããšãã§ããã
ãããã€ã¡ã³ãã§OpenVINO ãšã¯ã¹ããŒãã䜿çšãã
OpenVINO ãã¡ã€ã«ãå ¥æããããOpenVINO Runtime ã䜿çšããŠã¢ãã«ãå®è¡ã§ããŸããã©ã³ã¿ã€ã ã¯ããµããŒããããŠãããã¹ãŠã®ã€ã³ãã«ã»ããŒããŠã§ã¢ã§æšè«ãè¡ãããã®çµ±äžããã API ãæäŸããŸãããŸããã€ã³ãã«Â® ã㌠ããŠã§ã¢éã®ããŒããã©ã³ã·ã³ã°ãéåæå®è¡ãªã©ã®é«åºŠãªæ©èœãæäŸããŸããæšè«ã®å®è¡ã«é¢ãã詳现ã¯ããInference withOpenVINO Runtime Guide ( ã©ã³ã¿ã€ã ã»ã¬ã€ã)ããåç §ããŠãã ããã
Runtimeã§ã¢ãã«ãæ£ããã»ããã¢ããããŠäœ¿çšããã«ã¯ãXMLãã¡ã€ã«ãšBINãã¡ã€ã«ãããã³å ¥åãµã€ãºãæ£èŠåã®ããã®ã¹ã±ãŒã«ãã¡ã¯ã¿ãŒãªã©ãã¢ããªã±ãŒã·ã§ã³åºæã®èšå®ãå¿ èŠã§ããããšãå¿ããªãã§ãã ããã
ãããã€ã¡ã³ãã»ã¢ããªã±ãŒã·ã§ã³ã§ã¯ãéåžžã以äžã®æé ãå®è¡ããïŒ
- ãäœæããŠOpenVINO ãåæåããã
core = Core()
. - ã¢ãã«ãããŒãããã«ã¯
core.read_model()
ã¡ãœããã䜿çšããã - ã¢ãã«ãã³ã³ãã€ã«ããã«ã¯
core.compile_model()
é¢æ°ã§ããã - å ¥åïŒç»åãããã¹ããé³å£°ãªã©ïŒãæºåããã
- ã䜿çšããŠæšè«ãå®è¡ããã
compiled_model(input_data)
.
ãã詳现ãªæé ãšã³ãŒãã»ã¹ããããã«ã€ããŠã¯ãOpenVINO ããã¥ã¡ã³ããŸãã¯API ãã¥ãŒããªã¢ã«ãåç §ããŠãã ããã
OpenVINO YOLOv8 ãã³ãããŒã¯
YOLOv8 PyTorch ã ã ã ã®4ã€ã®ç°ãªãã¢ãã«ãã©ãŒãããã§ã ããŒã ã«ãã£ãŠå®è¡ãããŸããããã³ãããŒã¯ã¯Intel Flex GPUãšArc GPUãããã³Intel Xeon CPUã§FP32粟床ã§å®è¡ãããïŒFP32粟床ã§ã¯TorchScript ONNX OpenVINO Ultralytics half=False
åŒæ°ïŒã
泚
以äžã®ãã³ãããŒã¯çµæã¯åèå€ã§ãããã·ã¹ãã ã®æ£ç¢ºãªããŒããŠã§ã¢ããã³ãœãããŠã§ã¢æ§æããã³ãããŒã¯å®è¡æã®ã·ã¹ãã ã®çŸåšã®äœæ¥è² è·ã«ãã£ãŠç°ãªãå ŽåããããŸãã
ãã¹ãŠã®ãã³ãããŒã¯ã¯ openvino
Python ããã±ãŒãžããŒãžã§ã³ 2023.0.1.
ã€ã³ãã«ã»ãã¬ãã¯ã¹GPU
ã€ã³ãã«Â® ããŒã¿ã»ã³ã¿ãŒ GPU Flex ã·ãªãŒãºã¯ãã€ã³ããªãžã§ã³ããªããžã¥ã¢ã«ã¯ã©ãŠãåãã«èšèšãããæ±çšæ§ã®é«ãå ç¢ãªãœãªã¥ãŒã·ã§ã³ã§ãããã®GPUã¯ãã¡ãã£ã¢ã»ã¹ããªãŒãã³ã°ãã¯ã©ãŠãã»ã²ãŒãã³ã°ãAIããžã¥ã¢ã«æšè«ãä»®æ³ãã¹ã¯ãããã»ã€ã³ãã©ã¹ãã©ã¯ãã£ãŒã®ã¯ãŒã¯ããŒããªã©ãå¹ åºãã¯ãŒã¯ããŒãããµããŒãããŸãããã®GPUã¯ããã®ãªãŒãã³ã»ã¢ãŒããã¯ãã£ãšAV1ãšã³ã³ãŒãã®ãã«ãã€ã³ã»ãµããŒãã§éç«ã£ãŠãããé«æ§èœã§ã¯ãã¹ã¢ãŒããã¯ãã£ã®ã¢ããªã±ãŒã·ã§ã³ã«æšæºããŒã¹ã®ãœãããŠã§ã¢ã»ã¹ã¿ãã¯ãæäŸããŸããFlexã·ãªãŒãºGPUã¯å¯åºŠãšå質ã«æé©åãããŠãããé«ãä¿¡é Œæ§ãå¯çšæ§ãæ¡åŒµæ§ãæäŸããŸãã
以äžã®ãã³ãããŒã¯ã¯ãã€ã³ãã«Â® ããŒã¿ã»ã³ã¿ãŒ GPU Flex 170 㧠FP32 粟床ã§å®è¡ãããŠããŸãã
ã¢ãã« | ãã©ãŒããã | ã¹ããŒã¿ã¹ | ãµã€ãº (MB) | mAP50-95(B) | æšè«æéïŒms/imïŒ |
---|---|---|---|---|---|
YOLOv8n | PyTorch | â | 6.2 | 0.3709 | 21.79 |
YOLOv8n | TorchScript | â | 12.4 | 0.3704 | 23.24 |
YOLOv8n | ONNX | â | 12.2 | 0.3704 | 37.22 |
YOLOv8n | OpenVINO | â | 12.3 | 0.3703 | 3.29 |
YOLOv8s | PyTorch | â | 21.5 | 0.4471 | 31.89 |
YOLOv8s | TorchScript | â | 42.9 | 0.4472 | 32.71 |
YOLOv8s | ONNX | â | 42.8 | 0.4472 | 43.42 |
YOLOv8s | OpenVINO | â | 42.9 | 0.4470 | 3.92 |
YOLOv8m | PyTorch | â | 49.7 | 0.5013 | 50.75 |
YOLOv8m | TorchScript | â | 99.2 | 0.4999 | 47.90 |
YOLOv8m | ONNX | â | 99.0 | 0.4999 | 63.16 |
YOLOv8m | OpenVINO | â | 49.8 | 0.4997 | 7.11 |
YOLOv8l | PyTorch | â | 83.7 | 0.5293 | 77.45 |
YOLOv8l | TorchScript | â | 167.2 | 0.5268 | 85.71 |
YOLOv8l | ONNX | â | 166.8 | 0.5268 | 88.94 |
YOLOv8l | OpenVINO | â | 167.0 | 0.5264 | 9.37 |
YOLOv8x | PyTorch | â | 130.5 | 0.5404 | 100.09 |
YOLOv8x | TorchScript | â | 260.7 | 0.5371 | 114.64 |
YOLOv8x | ONNX | â | 260.4 | 0.5371 | 110.32 |
YOLOv8x | OpenVINO | â | 260.6 | 0.5367 | 15.02 |
ãã®è¡šã¯ã5ã€ã®ç°ãªãã¢ãã«(YOLOv8n,YOLOv8s,YOLOv8m,YOLOv8l,YOLOv8x)ãš4ã€ã®ç°ãªããã©ãŒããã(PyTorch,TorchScript,ONNX,OpenVINO)ã®ãã³ãããŒã¯çµæã§ãããåçµã¿åããã®ã¹ããŒã¿ã¹ããµã€ãºãmAP50-95(B)ã¡ããªãã¯ãæšè«æéã瀺ããŠããã
ã€ã³ãã«ã»ã¢ãŒã¯GPU
ã€ã³ãã«Â® Arcâ¢ã¯ãã€ã³ãã«ãå°çšGPUåžå Žã«åå ¥ããããšã象城ããŠããŸããArcâ¢ã·ãªãŒãºã¯ãAMDãNvidiaã®ãããªå€§æGPUã¡ãŒã«ãŒãšç«¶åããããã«èšèšãããŠãããããŒãããœã³ã³åžå Žãšãã¹ã¯ãããããœã³ã³åžå Žã®äž¡æ¹ã«å¯Ÿå¿ããŠããŸãããã®ã·ãªãŒãºã«ã¯ãããŒãããœã³ã³ã®ãããªã³ã³ãã¯ããªããã€ã¹åãã®ã¢ãã€ã«ã»ããŒãžã§ã³ãšããã¹ã¯ãããã»ã³ã³ãã¥ãŒã¿ãŒåãã®å€§åã§ãã匷åãªããŒãžã§ã³ããããŸãã
Arcâ¢ã·ãªãŒãºã¯3ã€ã®ã«ããŽãªãŒã«åãããŠããïŒArc⢠3ãArc⢠5ãArc⢠7ã®3ã€ã®ã«ããŽãªãŒã«åãããŠãããããããã®æ°åã¯æ§èœã¬ãã«ã瀺ããŠããŸããåã«ããŽãªãŒã«ã¯è€æ°ã®ã¢ãã«ããããGPUã¢ãã«åã®ãMãã¯ã¢ãã€ã«ãçµ±ååããªãšãŒã·ã§ã³ãæå³ããŸãã
åæã®ã¬ãã¥ãŒã§ã¯ãArcâ¢ã·ãªãŒãºãç¹ã«å èµA770M GPUã®çŽ æŽãããã°ã©ãã£ãã¯æ§èœãé«ãè©äŸ¡ãããŠããŸããArcâ¢ã·ãªãŒãºã®çºå£²ã¯å°åã«ãã£ãŠç°ãªããè¿æ¥äžã«è¿œå ã¢ãã«ãçºå£²ãããäºå®ã§ããã€ã³ãã«Â® Arc⢠GPUã¯ãã²ãŒã ããã³ã³ãã³ãå¶äœãŸã§ãããŸããŸãªã³ã³ãã¥ãŒãã£ã³ã°ã»ããŒãºã«å¯Ÿå¿ããé«æ§èœãœãªã¥ãŒã·ã§ã³ãæäŸããŸãã
以äžã®ãã³ãããŒã¯ã¯ãã€ã³ãã«Â® Arc 770 GPUã§FP32粟床ã§å®è¡ãããŠããŸãã
ã¢ãã« | ãã©ãŒããã | ã¹ããŒã¿ã¹ | ãµã€ãº (MB) | ã¡ããªãã¯/mAP50-95(B) | æšè«æéïŒms/imïŒ |
---|---|---|---|---|---|
YOLOv8n | PyTorch | â | 6.2 | 0.3709 | 88.79 |
YOLOv8n | TorchScript | â | 12.4 | 0.3704 | 102.66 |
YOLOv8n | ONNX | â | 12.2 | 0.3704 | 57.98 |
YOLOv8n | OpenVINO | â | 12.3 | 0.3703 | 8.52 |
YOLOv8s | PyTorch | â | 21.5 | 0.4471 | 189.83 |
YOLOv8s | TorchScript | â | 42.9 | 0.4472 | 227.58 |
YOLOv8s | ONNX | â | 42.7 | 0.4472 | 142.03 |
YOLOv8s | OpenVINO | â | 42.9 | 0.4469 | 9.19 |
YOLOv8m | PyTorch | â | 49.7 | 0.5013 | 411.64 |
YOLOv8m | TorchScript | â | 99.2 | 0.4999 | 517.12 |
YOLOv8m | ONNX | â | 98.9 | 0.4999 | 298.68 |
YOLOv8m | OpenVINO | â | 99.1 | 0.4996 | 12.55 |
YOLOv8l | PyTorch | â | 83.7 | 0.5293 | 725.73 |
YOLOv8l | TorchScript | â | 167.1 | 0.5268 | 892.83 |
YOLOv8l | ONNX | â | 166.8 | 0.5268 | 576.11 |
YOLOv8l | OpenVINO | â | 167.0 | 0.5262 | 17.62 |
YOLOv8x | PyTorch | â | 130.5 | 0.5404 | 988.92 |
YOLOv8x | TorchScript | â | 260.7 | 0.5371 | 1186.42 |
YOLOv8x | ONNX | â | 260.4 | 0.5371 | 768.90 |
YOLOv8x | OpenVINO | â | 260.6 | 0.5367 | 19 |
ã€ã³ãã«Xeon CPU
ã€ã³ãã«Â® Xeon® CPUã¯ãè€éã§èŠæ±ã®å³ããã¯ãŒã¯ããŒãåãã«èšèšãããé«æ§èœãªãµãŒããŒã°ã¬ãŒãã®ããã»ããµãŒã§ãããã€ãšã³ãã®ã¯ã©ãŠãã³ã³ãã¥ãŒãã£ã³ã°ãä»®æ³åãã人工ç¥èœãæ©æ¢°åŠç¿ã¢ããªã±ãŒã·ã§ã³ãŸã§ãXeon® CPUã¯ä»æ¥ã®ããŒã¿ã»ã³ã¿ãŒã«å¿ èŠãªãã¯ãŒãä¿¡é Œæ§ãæè»æ§ãæäŸããŸãã
ç¹çãã¹ãã¯ãXeon® CPUãé«ãæŒç®å¯åºŠãšã¹ã±ãŒã©ããªãã£ãå®çŸããäžå°äŒæ¥ãã倧äŒæ¥ãŸã§çæ³çãªç°å¢ãæäŸããããšã§ããã€ã³ãã«Â® Xeon® CPUãéžæããããšã§ãäŒæ¥ã¯ãè²»çšå¯Ÿå¹æãšéçšå¹çãç¶æããªãããæãèŠæ±ã®å³ããã³ã³ãã¥ãŒãã£ã³ã°ã»ã¿ã¹ã¯ãèªä¿¡ãæã£ãŠåŠçããã€ãããŒã·ã§ã³ãä¿é²ããããšãã§ããŸãã
以äžã®ãã³ãããŒã¯ã¯ã第4äžä»£ã€ã³ãã«Â® Xeon® Scalable CPUã§FP32粟床ã§å®è¡ãããŠããŸãã
ã¢ãã« | ãã©ãŒããã | ã¹ããŒã¿ã¹ | ãµã€ãº (MB) | ã¡ããªãã¯/mAP50-95(B) | æšè«æéïŒms/imïŒ |
---|---|---|---|---|---|
YOLOv8n | PyTorch | â | 6.2 | 0.3709 | 24.36 |
YOLOv8n | TorchScript | â | 12.4 | 0.3704 | 23.93 |
YOLOv8n | ONNX | â | 12.2 | 0.3704 | 39.86 |
YOLOv8n | OpenVINO | â | 12.3 | 0.3704 | 11.34 |
YOLOv8s | PyTorch | â | 21.5 | 0.4471 | 33.77 |
YOLOv8s | TorchScript | â | 42.9 | 0.4472 | 34.84 |
YOLOv8s | ONNX | â | 42.8 | 0.4472 | 43.23 |
YOLOv8s | OpenVINO | â | 42.9 | 0.4471 | 13.86 |
YOLOv8m | PyTorch | â | 49.7 | 0.5013 | 53.91 |
YOLOv8m | TorchScript | â | 99.2 | 0.4999 | 53.51 |
YOLOv8m | ONNX | â | 99.0 | 0.4999 | 64.16 |
YOLOv8m | OpenVINO | â | 99.1 | 0.4996 | 28.79 |
YOLOv8l | PyTorch | â | 83.7 | 0.5293 | 75.78 |
YOLOv8l | TorchScript | â | 167.2 | 0.5268 | 79.13 |
YOLOv8l | ONNX | â | 166.8 | 0.5268 | 88.45 |
YOLOv8l | OpenVINO | â | 167.0 | 0.5263 | 56.23 |
YOLOv8x | PyTorch | â | 130.5 | 0.5404 | 96.60 |
YOLOv8x | TorchScript | â | 260.7 | 0.5371 | 114.28 |
YOLOv8x | ONNX | â | 260.4 | 0.5371 | 111.02 |
YOLOv8x | OpenVINO | â | 260.6 | 0.5371 | 83.28 |
ã€ã³ãã«ã»ã³ã¢CPU
ã€ã³ãã«Â® Core® ã·ãªãŒãºã¯ãã€ã³ãã«ã«ããé«æ§èœããã»ããµãŒã®ã·ãªãŒãºã§ãããã©ã€ã³ãããã«ã¯ãCore i3ïŒãšã³ããªãŒã¬ãã«ïŒãCore i5ïŒãããã¬ã³ãžïŒãCore i7ïŒãã€ãšã³ãïŒãCore i9ïŒãšã¯ã¹ããªãŒã ããã©ãŒãã³ã¹ïŒããããŸããåã·ãªãŒãºã¯ãæ¥åžžçãªã¿ã¹ã¯ããèŠæ±ã®å³ãããããã§ãã·ã§ãã«ãªã¯ãŒã¯ããŒããŸã§ãããŸããŸãªã³ã³ãã¥ãŒãã£ã³ã°ã»ããŒãºãšäºç®ã«å¯Ÿå¿ããŠããŸããæ°ããäžä»£ã«ãªãããšã«ãããã©ãŒãã³ã¹ããšãã«ã®ãŒå¹çãæ©èœãæ¹åãããŠããŸãã
以äžã®ãã³ãããŒã¯ã¯ã第13äžä»£ã€ã³ãã«Â® Core® i7-13700H CPUã䜿çšããFP32粟床ã§å®è¡ãããŠããŸãã
ã¢ãã« | ãã©ãŒããã | ã¹ããŒã¿ã¹ | ãµã€ãº (MB) | ã¡ããªãã¯/mAP50-95(B) | æšè«æéïŒms/imïŒ |
---|---|---|---|---|---|
YOLOv8n | PyTorch | â | 6.2 | 0.4478 | 104.61 |
YOLOv8n | TorchScript | â | 12.4 | 0.4525 | 112.39 |
YOLOv8n | ONNX | â | 12.2 | 0.4525 | 28.02 |
YOLOv8n | OpenVINO | â | 12.3 | 0.4504 | 23.53 |
YOLOv8s | PyTorch | â | 21.5 | 0.5885 | 194.83 |
YOLOv8s | TorchScript | â | 43.0 | 0.5962 | 202.01 |
YOLOv8s | ONNX | â | 42.8 | 0.5962 | 65.74 |
YOLOv8s | OpenVINO | â | 42.9 | 0.5966 | 38.66 |
YOLOv8m | PyTorch | â | 49.7 | 0.6101 | 355.23 |
YOLOv8m | TorchScript | â | 99.2 | 0.6120 | 424.78 |
YOLOv8m | ONNX | â | 99.0 | 0.6120 | 173.39 |
YOLOv8m | OpenVINO | â | 99.1 | 0.6091 | 69.80 |
YOLOv8l | PyTorch | â | 83.7 | 0.6591 | 593.00 |
YOLOv8l | TorchScript | â | 167.2 | 0.6580 | 697.54 |
YOLOv8l | ONNX | â | 166.8 | 0.6580 | 342.15 |
YOLOv8l | OpenVINO | â | 167.0 | 0.0708 | 117.69 |
YOLOv8x | PyTorch | â | 130.5 | 0.6651 | 804.65 |
YOLOv8x | TorchScript | â | 260.8 | 0.6650 | 921.46 |
YOLOv8x | ONNX | â | 260.4 | 0.6650 | 526.66 |
YOLOv8x | OpenVINO | â | 260.6 | 0.6619 | 158.73 |
çµæãåçŸãã
äžèšã®Ultralytics ã®ãã³ãããŒã¯ããã¹ãŠã®ãšã¯ã¹ããŒããã©ãŒãããã§åçŸããã«ã¯ã以äžã®ã³ãŒããå®è¡ããŠãã ããïŒ
äŸ
ãã³ãããŒã¯ã®çµæã¯ãã·ã¹ãã ã®æ£ç¢ºãªããŒããŠã§ã¢ãšãœãããŠã§ã¢ã®æ§æãããã³ãã³ãããŒã¯ã®å®è¡æã®ã·ã¹ãã ã®çŸåšã®äœæ¥è² è·ã«ãã£ãŠç°ãªãå¯èœæ§ãããããšã«æ³šæããŠãã ãããæãä¿¡é Œæ§ã®é«ãçµæãåŸãã«ã¯ãç»åæ°ã®å€ãããŒã¿ã»ããã䜿çšããŸãã data='coco128.yaml' (128 val images), or
data='coco.yaml'` (5000 val images)ã
çµè«
ãã³ãããŒã¯çµæã¯ãYOLOv8 ã¢ãã«ãOpenVINO 圢åŒã«ãšã¯ã¹ããŒãããããšã®å©ç¹ãæ確ã«ç€ºããŠãããæ§ã ãªã¢ãã«ãããŒããŠã§ã¢ãã©ãããã©ãŒã ã«ãããŠãOpenVINO ãã©ãŒãããã¯ãåçã®ç²ŸåºŠãç¶æããªãããæšè«é床ã®ç¹ã§äžè²«ããŠä»ã®ãã©ãŒããããäžåã£ãŠããŸãã
ã€ã³ãã«Â® Data Center GPU Flexã·ãªãŒãºã§ã¯ãOpenVINO ãã©ãŒãããã¯ããªãªãžãã«ã®PyTorch ãã©ãŒãããã«æ¯ã¹ãŠæšè«é床ãçŽ10åéããªããŸãããXeon CPUã§ã¯ãOpenVINO ãã©ãŒãããã¯PyTorch ãã©ãŒãããã®2åã®é床ã§ãããã¢ãã«ã®ç²ŸåºŠã¯ãç°ãªããã©ãŒãããéã§ã»ãŒåãã§ããã
ãã³ãããŒã¯ã¯ããã£ãŒãã©ãŒãã³ã°ã¢ãã«ãå±éããããŒã«ãšããŠã®OpenVINO ã®æå¹æ§ã匷調ããŠããŸããã¢ãã«ãOpenVINO 圢åŒã«å€æããããšã§ãéçºè ã¯å€§å¹ ãªæ§èœåäžãéæããããšãã§ãããããã®ã¢ãã«ãå®äžçã®ã¢ããªã±ãŒã·ã§ã³ã«å±éããããšã容æã«ãªããŸãã
OpenVINO ã®è©³ããæ å ±ã䜿çšæ¹æ³ã«ã€ããŠã¯ã OpenVINO ã®å ¬åŒããã¥ã¡ã³ããåç §ããŠãã ããã
äœææ¥ïŒ2023-11-12 æŽæ°æ¥ïŒ2024-04-18
èè ïŒGlenn-Jocher(9),Abirami-Vina(1),RizwanMunawar(1),Burhan-Q(1)