All notable changes to this project will be documented in this file.
imageMarkupFaces
set toTrue
orFalse
to indicate showing bounding boxes on detected faces (overridesimageMarkup
value)imageMarkupObjects
set toTrue
orFalse
to indicate showing bounding boxes on detected objects (overridesimageMarkup
value)imageMarkupMotion
set toTrue
orFalse
to indicate showing bounding boxes on detected motion (overridesimageMarkup
value)
- Corrected default value for
objModelType
set tossd
for direct module execution
detector.analyze()
now supports overrides to core setting by supplying the boolean values for detectFaces, detectObjects, and/or detectMotion directly to this function.objDetectList
now enables you to specify a list of objects to look for (must be one of the values in the object labels list).- Enabled
--camera-device
for cli calls to specify what device or RTSP stream to use as the source. objModelType
now enablesyolo
orssd
options for model usage. (Yolo leverages yolov7-tiny)
- Default object detection model changed from MobileNet V3 SSD to Yolov7
- Fixed spelling mistakes in documentation for option variables
- Optimized hierarchy for image processing to improve overall performance and lag
- Corrected error in
--no-faces
option that was preventing it from taking affect - Corrected error in bounding boxes when scale factor is larger than 1.0 (which is not recommended)
- Corrected bad local import in main init()
- Disabled optimization for processing individual faces rather than entire image due to multi-channel image issues
orientation
option with values of 0, 90, 180, or 270 to rotate the input image by the desired degree.
- Changed dependency from opencv-python to opencv-contrib-python (Raspberry Pi requires this for Face/Object detection)
- Cleaned up mistake in Detector documentation with invalid object name in example
- Fixed but in rt_secs/rtSets for runtime stats.
- Multiple updates to documentation
- Corrected a bug when setting font and outline colors using tuples for running as module
This is the initial version and has not be heavily validated or tested. Stable versions should be available in v1.0 or later.
- CLI Parameters to enable calling module directly (python -m kenzy_image)
- Default to enable all detections (Face, Object, and Motion)
- Arguments added to enable faces to be recognized e.g.
--faces /path/image1.jpg LNXUSR1
(Can add multiple--faces
arguments.) - Added
--config
to enable specifying a JSON formatted file with all options configured. (See docs for more info.) - Use
--help
for full option list
- Renamed module to kenzy_image