Hi,
Currently I am working on a deepstream segmentation application for industrial defect.
As I can understand, the current Segmentation model is in .engine format of TensorRT.
I want to use this .engine model as the pre-trained model and then port that into the keras program so that I can continue perform the segmentation result accuray analysis (IoU, plot, etc.)
Can we change the tensorRT model in .engine format into .h5 format of Keras/Tensorflow?
Thanks