For those with no camera

vamarc · May 11, 2020, 11:54am

Hello All,

I was a bit disappointed as all my 3 USB web cameras did not work with Nano.

They display fine with fswebcam app (but not with Cheese).
I did not want to spend more money to buy a new one.
Anyways… It was hard for me to find a way to use videos files run samples. After some digging I finally put together the following code:

github.com

mvaranda/jetson-inference/blob/master/examples/mv_detection.py

import jetson.inference
import jetson.utils
import os, sys
import cv2

def app(filename):
  net = jetson.inference.detectNet("ssd-mobilenet-v2", threshold=0.5)
  display = jetson.utils.glDisplay()
  t = "i"
  ext = os.path.splitext(filename)[1]
  ext = ext.lower()
  if ext == "":           # Image is from Camera
    print("Image from camera")
    camera = jetson.utils.gstCamera()
    t = "c"
  elif ext == ".mp4" or ext == ".mov" or ext == ".avi":     # Image is from video file
    print("image from video file")
    t = "v"
  else:                   # Image is from a image file
    print("video from image file")

This file has been truncated. show original

Run run use the command: “python3 mv_detection.py <your_video_file>”

Know bug: If you click over the window while playing the program hangs. Need to kill the process in order to terminate/close it.

We may use this topic to join cameraless samples.

vamarc · May 11, 2020, 11:58am

Note that in my initial post for some reason the code is missing “This file has been truncated” (not sure why is truncated). Click “show original” to see the entire code.

DaneLLL · May 12, 2020, 1:13am

Hi,
Please share brand and model ID of the cameras. Not sure but it sounds like it should be run as RTSP sources. With more detail we can check and advise.