Pose Estimation example on i.MX8M Plus results in an error

[aekam_parmar]

Hello,

I am exploring ML capabilities of i.MX8M Plus and experienced some problems.

When I run this pipeline, that has TFLite converted and int8 quantized pose estimation model, it results in error.

gst-launch-1.0 filesrc location=../downloads/media/movies/Conditioning_Drill_1-_Power_Jump.webm.480p.vp9.webm ! matroskademux ! vpudec !  videocrop left=-1 right=-1 top=-1 bottom=-1 ! video/x-raw,width=480,height=480 ! imxvideoconvert_g2d videocrop-meta-enable=true ! video/x-raw,width=480,height=480,format=RGBA ! tee name=t t. ! queue name=thread-nn max-size-buffers=2 leaky=2 ! imxvideoconvert_g2d ! video/x-raw,width=192,height=192,format=RGBA ! videoconvert ! video/x-raw,format=RGB ! tensor_converter ! tensor_transform mode=typecast option=uint8 ! tensor_filter framework=tensorflow-lite model=../downloads/models/pose/movenet_single_pose_lightning.tflite custom=Delegate:External,ExtDelegateLib:libvx_delegate.so ! tensor_sink name=tensor_sink t. ! queue name=thread-img max-size-buffers=2 ! videoconvert ! cairooverlay name=cairooverlay ! autovideosink
** Message: 08:37:35.801: accl = cpu
Vx delegate: allowed_cache_mode set to 0.
Vx delegate: allowed_builtin_code set to 0.
Vx delegate: error_during_init set to 0.
Vx delegate: error_during_prepare set to 0.
Vx delegate: error_during_invoke set to 0.
ERROR: Fallback unsupported op 53 to TfLite
ERROR: Fallback unsupported op 53 to TfLite
ERROR: Fallback unsupported op 53 to TfLite
ERROR: Fallback unsupported op 88 to TfLite
ERROR: Fallback unsupported op 53 to TfLite
ERROR: Fallback unsupported op 53 to TfLite
ERROR: Fallback unsupported op 88 to TfLite
Setting pipeline to PAUSED ...
====== VPUDEC: 4.7.0 build on May 20 2022 06:51:21. ======
	wrapper: 3.0.0 (VPUWRAPPER_ARM64_LINUX Build on Jun  2 2022 08:51:59)
	vpulib: 1.1.1
	firmware: 1.1.1.43691
Pipeline is PREROLLING ...
Redistribute latency...
W [HandleLayoutInfer:268]Op 162: default layout inference pass.
W [HandleLayoutInfer:268]Op 74: default layout inference pass.
W [HandleLayoutInfer:268]Op 162: default layout inference pass.
W [HandleLayoutInfer:268]Op 162: default layout inference pass.
ERROR: Failed to verify graph
ERROR: Node number 157 (Vx Delegate) failed to invoke.

** (gst-launch-1.0:19436): CRITICAL **: 08:37:40.262: Failed to invoke

** (gst-launch-1.0:19436): CRITICAL **: 08:37:40.263: /usr/lib/libnnstreamer-single.so(_backtrace_to_string+0x40) [0xffff740173c0]/usr/lib/gstreamer-1.0/libnnstreamer.so(+0x3d9ac) [0xffff6e0fd9ac]/usr/lib/libgstbase-1.0.so.0(+0x48424) [0xffff87788424]/usr/lib/libgstbase-1.0.so.0(+0x47af4) [0xffff87787af4]/usr/lib/libgstreamer-1.0.so.0(+0x8e6d8) [0xffff87dce6d8]/usr/lib/libgstreamer-1.0.so.0(+0x90568) [0xffff87dd0568]/usr/lib/libgstbase-1.0.so.0(+0x47c08) [0xffff87787c08]/usr/lib/libgstreamer-1.0.so.0(+0x8e6d8) [0xffff87dce6d8]/usr/lib/libgstreamer-1.0.so.0(+0x90568) [0xffff87dd0568]/usr/lib/gstreamer-1.0/libnnstreamer.so(+0x1fba0) [0xffff6e0dfba0]/usr/lib/libgstreamer-1.0.so.0(+0x8e6d8) [0xffff87dce6d8]/usr/lib/libgstreamer-1.0.so.0(+0x90568) [0xffff87dd0568]/usr/lib/libgstbase-1.0.so.0(+0x47c08) [0xffff87787c08]/usr/lib/libgstreamer-1.0.so.0(+0x8e6d8) [0xffff87dce6d8]/usr/lib/libgstreamer-1.0.so.0(+0x90568) [0xffff87dd0568]/usr/lib/libgstbase-1.0.so.0(+0x47c08) [0xffff87787c08]/usr/lib/libgstreamer-1.0.so.0(+0x8e6d8) [0xffff87dce6d8]/usr/lib/libgstreamer-1.0.so.0(+0x90568) [0xffff87dd0568]/usr/lib/libgstbase-1.0.so.0(+0x47c08) [0xffff87787c08]/usr/lib/libgstreamer-1.0.so.0(+0x8e6d8) [0xffff87dce6d8]


** (gst-launch-1.0:19436): CRITICAL **: 08:37:40.263: Calling invoke function (inference instance) of the tensor-filter subplugin (tensorflow-lite for ../downloads/models/pose/movenet_single_pose_lightning.tflite) has failed with error code (-1).

^Chandling interrupt. (0.0 %)
Interrupt: Stopping pipeline ...
Setting pipeline to NULL ...
Freeing pipeline ...

I have observed other people facing the same problem but I could not reply on the topic for some technical reasons.

Has anyone found any solution to this?