OAID
运行Khadas下面的人脸检测和关键点检测出现问题
没有检测结果的输出提示如下： Tengine plugin allocator TIMVX is registered. vsiplugin reshape output tensor addr is 0x55a4c31d10 vsiplugin reshape output tensor addr is 0x55a4c31db0 vsiplugin reshape output tensor addr is 0x55a4c31e50 vsiplugin reshape output tensor addr is 0x55a4c31ef0 vsiplugin reshape output tensor addr is 0x55a4c32530 vsiplugin reshape output tensor addr is 0x55a4c320b0 vsiplugin reshape output tensor addr is 0x55a4c32150 vsiplugin reshape output tensor addr is 0x55a4c321f0 TENGINE_LAYOUT_NCHW goto RELEASE data size is 4472. data size is 4472. form : 3 ,to : 0 vsiplugin reshape output tensor addr is 0x55a5af9cc0 TENGINE_LAYOUT_NCHW goto RELEASE data size is 128. TENGINE_LAYOUT_NCHW goto RELEASE data size is 1724. init Over Detect Run Graph cost:1.22803 1 Run Over Tensor size mismatch 256 vs 512. /home/khadas/projects/khadas/vsiplugin-master/vxnpu/driver/vx_device.c:363 set input data for tensor 0 failed run subgraph 2 error! Landmark Run Graph cost:1.41992 Run Over ############################ 0，环境Ubuntu20.04，[ 1190.488197] Galcore version 6.4.3.p0.286725 1、采用最新的tengine编译的libtengine-lite.so，采用自带的tm_classification_timvx.c这个函数得到的可执行程序是可以使用npu的。 2、sudo apt install tengine-libs采用这种方式安装，解决了“load plugin failed: libvxplugin.so: cannot open shared object file: No such file or directory”
@Crusoekid 没有运行成功。人脸检测成功了，但是关键点检测出现问题，提示如下： 没有检测结果的输出提示如下： Tengine plugin allocator TIMVX is registered. vsiplugin reshape output tensor addr is 0x55a4c31d10 vsiplugin reshape output tensor addr is 0x55a4c31db0 vsiplugin reshape output tensor addr is 0x55a4c31e50 vsiplugin reshape output tensor addr is 0x55a4c31ef0 vsiplugin reshape output tensor addr is 0x55a4c32530 vsiplugin reshape output tensor addr is 0x55a4c320b0 vsiplugin reshape output tensor addr is 0x55a4c32150 vsiplugin reshape output tensor addr is 0x55a4c321f0 TENGINE_LAYOUT_NCHW goto RELEASE data size is 4472. data size is 4472. form : 3 ,to : 0 vsiplugin reshape output tensor addr is 0x55a5af9cc0 TENGINE_LAYOUT_NCHW goto RELEASE data size is 128. TENGINE_LAYOUT_NCHW goto RELEASE data size is 1724. init Over Detect Run Graph cost:1.22803 1 Run Over Tensor size mismatch 256 vs 512. /home/khadas/projects/khadas/vsiplugin-master/vxnpu/driver/vx_device.c:363 set input data for tensor 0 failed run subgraph 2 error! Landmark Run Graph cost:1.41992 Run Over
@Crusoekid 原始版本是6.4.0的，我重新加载了最新的6.4.3的。 [ 17.459173] galcore irq number is 36. [ 17.459175] Galcore version 6.4.0.229426 [616128.307545] galcore: no symbol version for module_layout [616128.307579] galcore: loading out-of-tree module taints kernel. [616128.322622] galcore irq number is 36. [616128.322631] Galcore version 6.4.3.p0.286725
