Eye Picker is a simple GUI tool writen in wxPython that provides a way to select landmark locations in images and saves the results to a comma separated file (CSV).
This was originally written because I could not find any good free tools for this task. The tool is a simple one window GUI that lets the user view and select landmark locations for a directory full of images. This is a common problem in face recognition which often requires left and right eye coordinates for each face.
The tool is very general. It will try to read any image with these extensions (case insensitive): PGM, PPM, BMP, JPG , JPEG, GIF, PNG, TIF, TIFF, or PGM. There is no limit to the number of points that can be selected. The source code is also avalible in the pyvision distribution so this tool can be modified to read other image formats or more complex directory structures.
|Download Eye Picker|
|(Mac OS X / Windows)|
|Cross platform source code is available with the pyvision distribution|
|or go directly to the source file: EyePicker.py (requires wxPython).|
Image Annotation on iOS
- Viper GT - This is a java ground truth annotation tool from the Language and Media Processing Laboratory at University of Maryland. It annotates videos or directories of images with points, bounding boxes and oriented rectangles, ellipses, and polygons and circles. Annotations are saved as xml files. Viper also includes a performance evaluation tool.
- MFA - A manual face alignment tool from Zou Wilman. The tool is used to select landmark points in JPEG images. The capabilities are probably available similar to Eye Picker. (Windows only)
- Image Region Labeling - A matlab tool for labeling image regions.
- LabelMe - A website that allows anyone to upload images and manually segment and label object. The website is also an open image database and allows the labeled images to be downloaded for research purposes.
- Human Annotation Tool - Tool from Berkeley that is specifically for annotating the pose of humans.
- Wasm - A active shape model training tool for use with the Stasm active shape model software.
If you would like to add another image annotation tool to this list please post details to the Pyvision Group.
Using Eye Picker
- After opening the application you will be asked to select a directory containing images. Select the appropriate directory and click Open.
- The main window will open with a list of images in the left panel.
- Click one of the image names to open that image in the right panel.
- Adding points to the image:
- Adding a point: click on the image.
- Move or adjust a point: drag an existing point to a new location.
- Clear the points and restart: reload the image (selecting another image and then the current image) and then and then click anywhere that is not currently a selected point.
- Save the locations to a CSV file by selecting Save from the File menu.
The keyboard can be used to scroll through images by pressing the up and down arrows.
Facial Landmark Datasets
This tool is FREE for all to use, however if you like the tool please consider donating your data for others to use. Collecting facial landmark data is often a difficult and labor intensive tasks. Here I have a short list of landmark data that I have collected myself or I found on the net. If you know of any related data that this list is missing, please post a message to the PyVision Google Group or otherwise let me know.
|Face Dataset||Num. Images||Num. Landmarks||Notes||Credit|
|FERET||150||25||Included With the CSU Eval System||David S. Bolme|
|CMU PIE (Expression)||75||20||Included with pyvision. Get the file here||David S. Bolme|
|FRGC||1000||13||500 Controlled and 500 Uncontrolled plus 100 annotations (including BioID) by a second person for use as control data.||David S. Bolme|
|BioID||1521||20||Included with the database||???|
|IMM face database||240||58||Mikkel B. Stegmann|
|MUCT||3755||76||Provide more diversity of lighting, age, and ethnicity||Stephen Milborrow|
|CMU PIE, AU Coded, AR||3805+||3 (Eyes and Nose)||Face Labels||Ralph Gross|
|Bosphorus||4666||24||3D Scans||Savran & Sankur|
|Multiple Datasets||*||*||Annotations for BioID, AR, and XM2VTS||FGNet|
|LFW||13233||2||Eye Coordinates for Labeled Faces in the Wild thread||Nikolay Degtyarev|
|Annotated Facial Landmarks in the Wild||25000||21||A large collection of annotated Flickr covering a variety of pose, expression, ethnicity, age, and gender.||Graz University|
This is an interesting overview of skull anatomy.