search

radeonwu / TensorRT_TX2_20FPS

Target to achieve 20 fps of vehicle/pedestrian detection(Fast R-CNN) on TX2 via TensorRT
36 stars 11 forks source link

SSD and YOLO with TensorRT #1

Open marvrez opened 6 years ago

marvrez commented 6 years ago

Hi, great job on setting up Faster-RCNN with TensorRT! Are you also planning on doing the same for YOLO and SSD? :) And also what fps are you currently getting when running sampleFasterRCNN?

radeonwu commented 6 years ago

hi, marvrez, What I have done so far for Faster R-CNN on TX2 is changing the input from ppm files into video file, while the TensorRT version of Faster-RCNN code is from samples under Jetson 3.1 package. Frame rate wise, with 720p video file, I am getting around 10 fps.

marvrez commented 6 years ago

Thank you for the answer @radeonwu , are you also going to accelerate SSD and YOLO with TensorRT as said in the README?