Object Detection Performance on Tesla T4 GPU TensorRT 8.2
810FPS (bs=1)YOLOv6-N
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| YOLOv6-NHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 810 | 1,323 | 1.2 | |
| YOLOv6-NHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 785 | 1,215 | 1.3 | |
| YOLOX-TinyHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 766 | 1,393 | 1.3 | |
| YOLOX-TinyHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 766 | 1,393 | 1.3 | |
| YOLOv5-NHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 702 | 843 | 1.4 | |
| YOLOv5-NHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 702 | 843 | 1.4 | |
| YOLOv6-THardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 469 | 677 | 2.1 | |
| YOLOv7-TinyHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 464 | 568 | 2.1 | |
| YOLOv7-TinyHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 464 | 568 | 2.1 | |
| YOLOv5-SHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 433 | 515 | 2.3 | |
| YOLOv5-SHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 433 | 515 | 2.3 | |
| YOLOv6-SHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 362 | 522 | 2.7 | |
| PPYOLOE-SHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 357 | 493 | 2.8 | |
| PPYOLOE-SHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 357 | 493 | 2.8 | |
| YOLOv6-SHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 345 | 498 | 2.9 | |
| YOLOX-SHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 313 | 489 | 2.6 | |
| YOLOX-SHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 313 | 489 | 2.6 | |
| YOLOv5-MHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 202 | 235 | 4.9 | |
| YOLOv5-MHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 202 | 235 | 4.9 | |
| YOLOv6-MHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 180 | 241 | 5.6 | |
| YOLOv6-MHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 178 | 238 | 5.6 | |
| PPYOLOE-MHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 163 | 210 | 6.1 | |
| PPYOLOE-MHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 163 | 210 | 6.1 | |
| YOLOX-MHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 159 | 204 | 5.3 | |
| YOLOX-MHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 159 | 204 | 5.3 | |
| YOLOv7Hardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 128 | 135 | 7.6 | |
| YOLOv7Hardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 128 | 135 | 7.6 | |
| YOLOv5-LHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 126 | 137 | 7.9 | |
| YOLOv5-LHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 126 | 137 | 7.9 | |
| YOLOv6-L-ReLUHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 111 | 145 | 9 | |
| PPYOLOE-LHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 110 | 145 | 9.1 | |
| PPYOLOE-LHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 110 | 145 | 9.1 | |
| YOLOv6-LHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 105 | 131 | 9.6 | |
| YOLOv6-LHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 105 | 125 | 9.5 | |
| YOLOX-LHardware=Tesla T4 GPU, Software=TensorRT 8.2, Batch Size=1 / 322022.09 | 104 | 117 | 9 | |
| YOLOX-LHardware=NVIDIA Tesla T4 GPU, TensorRT Version=8.2, Input Size=640x640, Precision=FP162023.01 | 104 | 117 | 9 |