BuenoCV-v0.0.0.11b

Platform: Windows 32bit

Description:
BuenoCV - is a tagging/detecting objects in an image.

Create virtual annotations in images, (you can move them around, create new ones, delete or resize existing ones). These are saved in CNTK https://www.microsoft.com/en-us/cognitive-toolkit/ structure, and can be used in the SDK examples, in couple of file formats:

• {original image name}.bboxes.labels.tsv = labels file for each one of the annotation generated to it.
• {original image name}.bboxes.tsv - the annotations file dimentions
• train_roi_file.txt - Concated annotations dimentions of all images in the given path
• train_img_file.txt - List of images file names to be used in the CNTK training

Create OpenCV classifiers --> and Create cropped images of positive, negative folders to be used in the classifiers

Run OpenCV Classifier Training --> It uses opencv https://opencv.org/ ANN algorithm in order to run a quick detection of the different objects, in a given new image. You can group each object in a cascade.xml file which will used later on for detection.

It uses opencv ANN algorithm in order to run a quick detection of the different objects, in a given new image. You can group each object in a cascade.xml file which will used later on for detection.

Cascade:
A classifier that represent an object, such as a face classifier, cat, fruit, etc.. Cascades are created by the training mechanism, and being used to detect object occurrence in a new picture.

Training:
Training is a procedure that takes positive and negative images, running on them a learning procedure, that eventually creates a cascade file.
Positive: an image where you can see the object
Negative: an image where you do not see the object
In order to run a proper training, you suppose to tag at least 50+ positive images, count of negative images will have to be two times bigger (or more) than the positive image count.

Features:

• Cascades: Ability to create a new cascade (By Pressing the Tool Buttons Bar '+' Button)
  • Create a cascade directory: with basic properties as name, size
  • Mark: Mark new positives and negatives with the NewMarker (being picked from the upper tool buttons bar ) -- Adding the NewMarker to the positive/negative list is done, by either pressing the Tool Buttons Bar Add icon, or by right clicking on the NewMarker rectangle.
  • Move: NewMarker can be moved around (dragged) via left Mouse Click and Mouse Move
  • Scale: NewMarker can be scaled in the correct cascade ratio -- Using SHIFT+MouseMove on the NewMarker
• Pick existing cascade definition from the upper right Cascades List
• Hide/Show all cascades detections ( Tool Buttons Bar Eye Button )
• Quick marking false positives: Fast Ability to mark false positives detections as negatives (by selecting the false positive rectangle and pressing Right Mouse Click -- remember to use the proper Tool Buttons Bar [positive/negative])
• Training: takes random marked negative images against each positive image. Final result of a training are cascades .xml files created for each group. While in the process, Four progress bars are presented logging the progress in the procedure.
• Detecting: You can load a directory that contains .jpg images, by picking a directory/filename on the upper editbox, and pressing ENTER -- on the left listbox panel you'll see all .jpgs from the picked directory -- if a cascade.xml detection find a match the app will mark a rectangle around it, in the cascade color defined in each cascade property.