search

jkjung-avt / yolov4_crowdhuman

A tutorial on training a DarkNet YOLOv4 model for the CrowdHuman dataset
MIT License
140 stars 49 forks source link

Can't detect my model #29

Open QuyNguyen87 opened 2 years ago

QuyNguyen87 commented 2 years ago

Hello, my model is traffic sign detection, I have downloaded my files cfg and weights and convert them to tensorRT as your guide, but it can't show the label name of traffic sign. Could you show me how can I fix it, thank you. Screenshot from 2022-04-08 12-49-45

jkjung-avt commented 2 years ago

https://github.com/jkjung-avt/tensorrt_demos/issues/421

QuyNguyen87 commented 2 years ago

I set up my model 416x416 in the cfg file and convert it to tensorrt but when I run it, the output was not 416x416 and the fps is only 2-2.5, are there anything I have to change in files? and how can I increase fps at least 10, thank you very much

jkjung-avt commented 2 years ago

If your cfg is 416x416, the created TensorRT YOLO engine should be 416x416. The TrtYOLO class would do inference in 416x416 and scale the bounding boxes back to coordinates in the original image.

The FPS depends heavily on the GPU (or Jetson platform) you are using. You should also consider using a light/faster model (such as YOLO-Tiny) if your primary goal is to achieve higher FPS (tradeoff against detection accuracy).