https://github.com/ky28059/stop-sign-detection

Feature-based stop sign detection?
https://github.com/ky28059/stop-sign-detection

Last synced: 3 months ago
JSON representation

Feature-based stop sign detection?

• Host: GitHub
• URL: https://github.com/ky28059/stop-sign-detection
• Owner: ky28059
• License: mit
• Created: 2024-07-31T01:48:44.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2024-07-31T01:52:41.000Z (11 months ago)
• Last Synced: 2024-07-31T04:31:58.386Z (11 months ago)
• Language: Python
• Size: 16.4 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# stop-sign-detection

To run:

- Place images containing stop signs in `./samples/positive` (some have already been provided).

- Place images *not* containing stop signs in `./samples/negative` (some have already been provided).

- Run `test.py` to iterate through each image and see what features the algorithm detects!

### About

The idea behind this project is loosely based around [this OpenCV guide about FLANN-based feature matching](https://docs.opencv.org/4.x/d5/d6f/tutorial_feature_flann_matcher.html),

using OpenCV's [ORB](https://docs.opencv.org/4.x/d1/d89/tutorial_py_orb.html) keypoint detector instead of the SURF

detection used in the guide for licensing reasons.

Do stop signs naturally lend themselves to feature tracking? Not *really*. But there are seemingly enough features in

the stop sign text to differentiate stop-sign photos from non stop-sign photos with decent accuracy.



     





     



### Alternatives to consider

- Using an OCR algorithm like [tesseract](https://pypi.org/project/pytesseract/) to detect the word "stop" in each image,

  though it will take a long time to run (probably insufficient for RACECAR purposes).

- Training a CNN on stop sign images; this will likely be faster due to the TPU on the racecars.