I am trying to modify the TorchVideo app to run video detection model instead of video classification.
I have converted a SlowFast model to TorchScript .ptl format with the help of pytorchvideo and by modifying the model conversion script available here.
Now the issue is, it seems that roi_align( ) function is missing in pytorch_android_torchvision:1.9.0 which is required for running detection models. The runtime error message from Android Studio is shown below
E/AndroidRuntime: FATAL EXCEPTION: main
Process: org.pytorch.demo.torchvideo, PID: 23380
java.lang.RuntimeException: Unable to start activity ComponentInfo{org.pytorch.demo.torchvideo/org.pytorch.demo.torchvideo.MainActivity}:
com.facebook.jni.CppException: Following ops cannot be found. Check fburl.com/missing_ops for the fix.
{torchvision::roi_align, } ()
Exception raised from print_unsupported_ops_and_throw at ../torch/csrc/jit/mobile/import.cpp:206
Hi @arunponnusamy, it appears that SlowFast detection model uses 2 input while the conversion script provided in this repo only accept 1 input. How did you make it work?
I am trying to modify the TorchVideo app to run video detection model instead of video classification.
I have converted a SlowFast model to TorchScript
.ptl
format with the help of pytorchvideo and by modifying the model conversion script available here.Now the issue is, it seems that
roi_align( )
function is missing inpytorch_android_torchvision:1.9.0
which is required for running detection models. The runtime error message from Android Studio is shown belowAny help on this is much appreciated.