A tool for image landmark annotation written in Python
merged.mp4
python image_landmarking.py <path to image folder> -nl <number of landmarks per image> -dw <display width>
example:
python image_landmarking.py images -dw 720
To save the result, quit the application by closing the window or pressing esc. The result will appear in the image folder as landmarks.pickle.
This tool uses Pygame and OpenCV to interactively place, move, remove and tag landmarks on a serie of images. Image keypoints are very useful in many computer vision application including face alignment.
The controls are displayed in the console when launching the application.
| Control | Action |
|---|---|
| right click | Place a landmark under the cursor |
| left click | Grab a landmark under the cursor |
| mouse wheel | zoom |
| wheel click and drag | translation |
| lctrl + left click | Falg landmark as occluded |
| backspace | remove last placed landmark |
| c | clear the landmarks on the current image |
| right arrow | next image |
| left arrow | previous image |
Landmarks are color coded to ease the process.
| Color | Meaning |
|---|---|
| blue | Default |
| magenta | Last placed |
| green or orange | Flagged as occluded |
- python 3.7+
- pygame
- opencv-python
- numpy
- click