Skip to content

Frankushima/pete

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Project P.E.T.E

Basically runs a version of yolov7 on two cameras that shows the camera input + yolov7 bounding boxes on a GUI checklist thing

Its also runs on a Jetson Orin Nano.

Note: a lot of stuff is hardcoded due to basically trial and error to get this stubborn ass board to run this thing

Setup for running with video (Read This):

  1. Download testing videos from Slack or use ur own
  2. Download model weight files from here
  3. Run a test run with default detect:
python detect.py --weights=yolov7-tiny.pt --source=test_videos/bottomBracketInstall.MOV --nosave --view-img --no-trace
  1. Run a command similar to this one (This is the actual command you will use from now on):
python GUI_detect.py --weights=yolov7-tiny.pt --source=test_videos/bottomBracketInstall.MOV --nosave --view-img --no-trace

Arguments:

  • weights = which model to use
  • source = test video file location
  • nosave = don't save results
  • view-img = needs to be there for the GUI to work
  • no-trace = skip tracing model step to make bootup faster

Quirks with this setup:

  1. Video plays at slower speed to match detection rate (HACK: made it skip every other frame)

TODO:

  • Make video skip frames to simulate real time enviroment (DONE)
  • Add loading gif (DONE)
  • Add manual revert step (DONE)
  • Add substeps with pictures for context (WIP)
  • Add sensor stuff

FOR LIVE ONLY

Instructions for live object detection with two cameras:

  • Launch raspi's gstreamer pipeline (1296x972 for max viewing angle)
raspivid -t 0 -h 972 -w 1296 -fps 25 -hf -b 2000000 -o - | gst-launch-1.0 -v fdsrc ! h264parse ! rtph264pay config-interval=1 pt=96 ! gdppay ! tcpserversink sync=0 host=192.168.0.248 port=5000
  • Launch this script:
python GUI_detect.py --weights=Demo_Only_B40.pt --source=test_videos/bottomBracketInstall.MOV --nosave --view-img --no-trace

Quirks with this setup:

  1. Raspi will freeze and crash for no reason sometimes (often happens after this program terminates) HACK: Stop gstreamer pipeline before script terminates SHTF: REISUB board thru micro USB + keyboard

  2. If you switch USB devices while the model is loading, it might freeze the board... edge device things SHTF: REISUB board thru micro USB + keyboard

  3. On-board camera seems to only run properly with V4L2 instead of gstreamer with OpenCV HACK: hardcoded OpenCV flags

PROGRESS REPORTS

4/13:

  • Fixed procedure list scaling issues
  • Fixed focus bug on reverting steps
  • Added loading gif
  • Fixed exiting window not exiting program bug
  • Added pictures to steps (WIP)
  • Fixed gitignore to make sure stuff is downloaded correctly

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 33