In the test, the batch size of model is 1. The CPU and GPU parameters are as follows:
| type | number | |
|---|---|---|
| CPU | Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz | 6 |
| GPU | Tesla T4 15109MiB | 1 |
When using Adlik for model inference, the PaddlePaddle model is deployed in OpenVINO type.
| model | using PaddlePaddle (ms) | using Adlik(ms) |
|---|---|---|
| PPocr_det | 105.82 | |
| PPocr_rec | 18.91 | |
| PPocr_cls | 96.02 | |
| PPresnet50 | 16.36 |
When using Adlik for model inference, the PaddlePaddle model is deployed in TensorRT type.
| model | using PaddlePaddle (ms) | using Adlik(ms) |
|---|---|---|
| PPocr_det | 33.51 | |
| PPocr_rec | 13.72 | |
| PPocr_cls | 1.56 | |
| PPresnet50 | 4.93 | |
| PPyolo_mobile_net | 4.20 |