Here are few steps required to get Imagenet based on OpenCV working DB845c ################## on HOST ################################### For some reason SNPE looks at /proc/cpuinfo to figure out the soc info, this patch should make it happy for DB845c http://people.linaro.org/~srinivas.kandagatla/0001-HACK-arm64-cpuinfo-Make-SNPE-happy-on-DB845c.patch Make sure that you have included this in your kernel before testing. ################# on TARGET DB845c ############################### Step 1: Build new fastrpc lib on DB845c $ git clone -b automake https://git.linaro.org/landing-teams/working/qualcomm/fastrpc.git $ cd fastrpc $ ./gitcompile $ make install Step 2: download Qualcomm Neural Processing SDK for AI version snpe-1.30.0.480 from https://developer.qualcomm.com/software/qualcomm-neural-processing-sdk and Unzip into /home/linaro folder as /home/linaro/snpe-1.30.0.480/ $ cp $ mount /dev/disk/by-partlabel/dsp_a /dsp $ cp /home/linaro/snpe-1.30.0.480/lib/dsp/* /dsp/cdsp/ Step 3: Start CDSP and userspace cdsprpcd deamon # start cdsp if its not already started. $ echo start >/sys/devices/platform/remoteproc-cdsp/remoteproc/remoteproc1/state #Run qrtr services if they have not already started $ qrtr-cfg 1 && qrtr-ns && rmtfs -v & #Mount dsp partition #export libcdsp path. $ export ADSP_LIBRARY_PATH=/dsp/cdsp #Start cdsp deamon $ cdsprpcd& Step 4: Install OpenCV on the Target $ sudo apt-get install libopencv-dev python3-opencv Step 5: Build SNPE Demo application and run ## To Build ## $ git clone https://git.linaro.org/people/srinivas.kandagatla/SNPEMobileNetCv.git $ cd SNPEMobileNetCv $ mkdir build $ cd build $ cmake .. $ make $ export LD_LIBRARY_PATH=/home/linaro/snpe-1.30.0.480/lib/aarch64-linux-gcc4.9/ $ ./SNPEMobileNetSSDCv This should show up a live video with MoblieNet SSD detections with boxes