Issue in loading FasterRCNN tlt engine model for prediction

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

hii @nvidias
I have converted my etlt frcnn model into .engine file Using TLT-converter.
I am facing issue in predictions. is there any python script through which I can load my frcnn engine file and pass image and see the detection results. for detectnet_v2 I was using peopling script from Nvidia forum . it was working for detectnet_v2 module but For FasterRCNN it is giving me binding issues.
Any help will be appreciable.
Thanks

2

Officially, TLT only provides tlt-infer for inference. End user can write their own scripts to run inference against trt engine.

For frcnn, its output is

  • an output of shape [batchSize, 1, keepTopK, 7] which contains nmsed box class IDs(1 value), nmsed box scores(1 value) and nmsed box locations(4 value)
  • another output of shape [batchSize, 1, 1, 1] which contains the output nmsed box count.
1 Like