TensorRt sampleUffFasterRcnn

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hi,

I wanna send inputs (‘activation_7/Relu’, ‘proposal’) to roi_pooling_conv_1/CropAndResize_new layer for some research purpose. How can I convert fasterrcnn.pb model with input node having two inputs [input_1, roi_pooling_conv_1/CropAndResize] to uff and to Tensorrt engine.

convert-to-uff command needs input node with (name, new_name, dtype, shape)

For input_1 I can give (input_1,input_1,float32,3,272,480). Similarly how can I specify inputs for roi_pooling_conv_1/CropAndResize_new ? and convert it to TensorRt engine

2

Hi,

I think current code is handling the same case “Send inputs (‘activation_7/Relu’, ‘proposal’) to roi_pooling_conv_1/CropAndResize_new layer”

Please refer to below link for more details:
https://github.com/NVIDIA/TensorRT/blob/master/samples/opensource/sampleUffFasterRCNN/config.py#L19

Thanks

3

I need to be able to add my own input proposal to ‘roi_pooling_conv_1/CropAndResize_new’ not the one calculated by the model.

4

Hi,
Please refer to below link for custom plugin layer:
https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/sampleUffFasterRCNN#sampleufffasterrcnn-plugins

Thanks