search

dusty-nv / jetson-inference

Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.
https://developer.nvidia.com/embedded/twodaystoademo
MIT License
7.72k stars 2.97k forks source link

how can I simultaneously run object detection and semantic segmentation with SegNet (street detection (fcn-resnet18-cityscapes)) through kammera stream #1587

Open hosseinAT opened 1 year ago

hosseinAT commented 1 year ago

I have a question, I have trained a new objects (wheelchair and wheelchair user) and the car should be able to recognize the wheelchair and wheelchair user when it enters the road.

In order to meet the criteria, one must develop a system that recognizes Wheelchair, Wheelchair_User and distinguishes it from other people. Here are some steps to achieve that:

  1. Image acquisition systems: Positioning the camera on the car roof, looking in the direction of travel, to monitor the road and capture images of people, Wheelchair_User and Wheelchair.
  2. Object Detection Algorithms: Using machine learning or computer vision algorithms to detect wheelchairs and people in the images.
  3. Street detection (fcn-resnet18-cityscapes): by which one can know whether the wheelchair and the wheelchair user are on the street or sidewalk (footpath).

how can I simultaneously run object detection and semantic segmentation with SegNet (street detection (fcn-resnet18-cityscapes)) through kammera stream? when i want to run both, i get an error saying i should stop one of them. can you please help me?

https://github.com/hosseinAT/Rollstuhl_und_Rollstuhlfahrererkennung_verwendung_auf_Kriterien.git

dusty-nv commented 1 year ago

Hi @hosseinAT, you should make your own Python or C++ program that uses both detectNet and segNet on the same camera image. I would pass overlay=none to detectNet.Detect() so that it doesn't overwrite the original image before processing it with segnet. Then after both the inference is complete, use detectNet.Overlay()

You can also see this example which does multiple DNNs - https://github.com/dusty-nv/jetson-inference/blob/master/docs/webrtc-flask.md

hosseinAT commented 1 year ago

Hello Dustin,

Thanks for the support and I tried for a few days but unfortunately it didn't work.

this is my training model: Rollstuhl_und_Rollstuhlfahrererkennung_verwendu... https://drive.google.com/file/d/1bHtQTiVNdGoO9WG1zLIDBNTHqFE_EdLu/view?usp=drive_web

Best regards

hossein

Am Fr., 24. März 2023 um 17:07 Uhr schrieb Dustin Franklin < @.***>:

Hi @hosseinAT https://github.com/hosseinAT, you should make your own Python or C++ program that uses both detectNet and segNet on the same camera image. I would pass overlay=none to detectNet.Detect() so that it doesn't overwrite the original image before processing it with segnet. Then after both the inference is complete, use detectNet.Overlay()

You can also see this example which does multiple DNNs - https://github.com/dusty-nv/jetson-inference/blob/master/docs/webrtc-flask.md

— Reply to this email directly, view it on GitHub https://github.com/dusty-nv/jetson-inference/issues/1587#issuecomment-1483053759, or unsubscribe https://github.com/notifications/unsubscribe-auth/A57ZBKSU2X5SXOYXTC4H45DW5XBDJANCNFSM6AAAAAAWGS5XX4 . You are receiving this because you were mentioned.Message ID: @.***>