Face_landmark_tflite mode inference in TRT 6 in jetson nano


I am using face_landmark model with 468 features for face recognition. Both these onnx files does not work throws some errors.

(Here is the link to the mediapipe model file: face_landmark.tflite, facenet_468_mediapipe.onnx (2.3 MB)
2nd repo: face_landmark.tflite (2.3 MB)
face_landmark_tflite.onnx (2.3 MB)
I could able to port it to TensorRT 8 (tested with ./trtexce command) and it works there without any errors.

But, I would like to do it in TensorRT 6. Is there a possibility to run it in TRT 6??

Any help would be really appreciated.
Thanks in advance.


Unfortunately, we don’t back-port the feature/implementation in TensorRT.
It’s recommended to use latest version for a better support.