Hi all,
In the concept of Tensorflow framework, If we generated the .pb file, we can easily run .pb model with importing of .pb file and then specific the inputs/outputs names in the last step run the model and this structural is general for all .pb models, I want to know, for running the .engine models(generated with TLT), Is there a right way like above for doing inference with TensorRT? If so, please put some python codes for this problem.
User can do inference with TRT engine.
See https://docs.nvidia.com/metropolis/TLT/tlt-getting-started-guide/index.html#gen_eng_tlt_converter
and https://docs.nvidia.com/metropolis/TLT/tlt-getting-started-guide/index.html#intg_model_deepstream
Output blob names:
For classification : predictions/Softmax
For DetectNet_v2: output_bbox/BiasAdd,output_cov/Sigmoid
For FasterRCNN: dense_class_td/Softmax,dense_regress_td/BiasAdd, proposal
For SSD, DSSD, RetinaNet: NMS
For YOLOv3: BatchedNMS
Input blob names:
For classification:input_1
For DetectNet_v2: input_1
For FasterRCNN: input_image
For SSD, DSSD, RetinaNet: Input
For YOLOv3: Input