I was successful in converting JDE tracking model to tensorRT as the input_shapes to model is clearn (1, 3, 608, 1088).
But for action model, feature extraction inputs to model is not clear, here boxes and objects are given as class objects, not tensor arrays and the size varies for each frame. So, how and where are these two class objects getting converted to arrays.
Any help/guidance with the conversion of AIA model to tensorRT is much appreciated
I was successful in converting JDE tracking model to tensorRT as the input_shapes to model is clearn (1, 3, 608, 1088).
But for action model, feature extraction inputs to model is not clear, here boxes and objects are given as class objects, not tensor arrays and the size varies for each frame. So, how and where are these two class objects getting converted to arrays.
Any help/guidance with the conversion of AIA model to tensorRT is much appreciated
Thanks