Table 2 The efficiency evaluation on three common smartphones with detailed configuration.
From: Deep learning-enabled mobile application for efficient and robust herb image recognition
Smartphone | HUAWEI Mate9 Pro | Xiaomi MI 9 | HUAWEI P30 Pro | |
|---|---|---|---|---|
CPU type | Kirin 960 | Qualcomm 855 | Kirin 980 + NPU | |
Basic frequency | 2.36 GHz | 2.83 GHz | 2.60 GHZ | |
RAM size | 4 GB | 8 GB | 8 GB | |
Android version | 9 | 9 | 10 | |
Screen resolution | 1080 x 1920 | 1080 x 2340 | 1080 x 2340 | |
Large DNN | Average running time | 342 ms | 255 ms | 187 ms |
Frames per second | 2.9 fps | 3.9 fps | 5.3 fps | |
Top1 accuracy | 70.61 ± 1.00 | |||
Top5 accuracy | 88.97 ± 0.41 | |||
Small DNN + transfer | Average running time | 130 ms | 73 ms | 59 ms |
Frames per second | 7.7 fps | 13.6 fps | 16.7 fps | |
Top1 accuracy | 70.97 ± 0.50 | |||
Top5 accuracy | 88.72 ± 0.79 | |||
Small DNN + transfer + cut | Average running time | 75 ms | 51 ms | 48 ms |
Frames per second | 13.4 fps | 19.6 fps | 21.0 fps | |
Top1 accuracy | 69.98 ± 1.09 | |||
Top5 accuracy | 87.71 ± 0.73 | |||
