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