基础 AI 推理示例
YOLOv5 目标检测（官方示例）
bash
# 进入C++示例目录
cd /usr/local/share/aidlite/examples/aidlite_qnn236/cpp
# 编译实例
mkdir build && cd build
cmake ..
make
# DSP加速运行（推荐，参数3对应.bin模型）
./qnn_yolov5_multi 3预期输出：
bash
current thread_idx[1] [9] get_output_tensor cost time : 0.955974
repeat [10] time , input[25.862552] --- invoke[22.108812] --- output[18.433402] --- sum[66.404766]ms
postprocess cost time : 0.178018 ms
Result id[0]-x1[472.021393]-y1[232.090714]-x2[561.091858]-y2[519.549500]
Verify result : idx[0] id[0] coverage_ratio[0.000000]
Verify result : idx[1] id[0] coverage_ratio[0.959014]
Result id[0]-x1[211.840363]-y1[246.274216]-x2[283.799805]-y2[514.897522]
Verify result : idx[0] id[0] coverage_ratio[0.927729]
Result id[0]-x1[114.483437]-y1[230.784622]-x2[201.432831]-y2[546.368408]
Verify result : idx[0] id[0] coverage_ratio[0.000000]
Verify result : idx[1] id[0] coverage_ratio[0.000000]
Verify result : idx[2] id[0] coverage_ratio[0.701955]
Result id[5]-x1[86.659409]-y1[124.519958]-x2[557.957275]-y2[478.171204]
Verify result : idx[0] id[5] coverage_ratio[0.919242]
The result display file[out_yolov5_qnn_trd1.jpg] for the example is located in the current working path
The inference result box_count[4]-verify_pass_count[3].Python 版 YOLOv5 推理
bash
# 进入python实例目录
cd /usr/local/share/aidlite/examples/aidlite_qnn236/python
python3 qnn_yolov5_multi.py 3预期输出：
bash
current [9] get_output_tensor cost time :0.0013930797576904297 ms
repeat [10] times , input[21.659135818481445]ms --- invoke[21.498918533325195]ms --- output[26.68595314025879]ms --- sum[69.84400749206543]ms
检测到4个区域
1 [472, 232, 561, 519] 0.8130854 person
2 [211, 246, 282, 514] 0.81079763 person
3 [114, 230, 200, 545] 0.7994484 person
4 [86, 124, 557, 477] 0.7512346 bus验证：查看生成的qnn_yolov5_multi.jpg，确认检测框位置正确。 