truth-and-crop

Convient GUI application for quickly ground-truthing semantic segmentation datasets in Python/OpenCV.

Dependencies

• python 3.4
• pyqt 4.x
• opencv 3.x
• numpy 1.11
• colorama 0.3
• scikit-image 0.12

To install PyQt4 with Anaconda, conda install pyqt=4

Usage

python truth_and_crop.py

Buttons

• Input File - Browse to image file to load.
• Output Path - Browse to root folder where output should be saved. Three subfolders are automatically created here.
• Previous/Next Image - If other images were found in same folder as Input File, you can jump between images with these buttons.
• Refresh - Discards changes.
• Crop - Switch between annotation mode and cropping mode.
• Toggle - Toggle annotations on and off to make it easier to see raw image. SLIC is only run on the image for the first toggle, subsequent toggles are much faster.
• Save - To write all cropped images and masks into appropriate subfolders under the path specified by 'Output Path'.