-
Add car tracking module using yolo3 models
- Tracking car system using yolov3 models
- Multithreding system
- multipe OCR options: pytesseract, easyocr, pyocr
- possible Arabic letter detection
- Possible multiple pattern of Plate models
- Video show without saving frame: skipping frames to speed up process
- Visualize:
- Blue box of detected car
- plate number and confidence (White/Black)
- yellow box if datetime exists in the screen
-
Separate frame processing of v0.1.1 and frame processing of v0.1.3 by when add --video argument
- if run_cil4sys without --video, v0.1.1 is executed
- if run_cil4sys --video, car tracking and license plate saving is proceeded (v0.1.3)
-
run_cil4sys could be run from any directory since cil4sys is installed
- default video: success detected license plate in 23s with 31/181 saved frames
-
Cleanup capturing Video process in video class
- write open_video_sources fucntion
- rewrite video_read, video_split and capture_video
- resolve conflict between Output_path and video_path
- upgrade opencv-python to 4.6.0.66: ultralytics 8.0.222 requires opencv-python>=4.6.0
- upgrade scipy to 1.8.1: issue with Ubuntu: undefined symbol: _PyGen_Send
-
Function Decomposition:
- Break down the main function into smaller functions to improve readability and maintainability. Each function should have a specific responsibility.
-
Logging:
- Instead of using print statements for informational messages, consider using Python's logging module. It provides more control over log messages.
-
Limit saving frames if new number plate is detected
- Use tracking system to detect motion
- Reduce motion detection to only car
- Do not save the save car/plate: skip frames
-
Use other ORC that seems to be more efficient
- Arabic option: try to capture arabic letters
- Use regex for multiple pattern like: standard tunisian plate, RS, Diplomatic...
-
Display video with results
- Box for ROI
- Plate number
- Confidence
-
Extract datetime during saving plate number
- Security video cam displays datetime in the upper-right corner of the screen
-
for recorded video:
- extract video sequence with only car/motion to improve performance