pixlabel  v1.0

A web-tool for rapid custom pixel labeling

Examine pixel color, count objects, or simply collect lots of ground-truth color data for regression and machine learning projects.







Instructions


Step 1
Load an image from your local hard drive or play with the example picture.


Step 2
Select a pixel array. A "Single pixel" means that information is recorded only for the pixel right underneath the crosshair cursor. Depending on the application, users might be interested in knowing the color value of neighboring pixels. So, you can also choose to record information in a 3x3, 5x5, and 7x7 pixel area. The central pixel is always the pixel under the crosshair cursor. The order in which pixels are recorded is illustrated with the 3x3 pixel area.


Step 3
Write a label, no need to press the Enter key. You can literally enter anything: Andromeda, flower, soil, elfs. Labels can have spaces. When you are done with the current label just write a new one and keep selecting pixels.


Step 4
Click on the areas of the image that represent the current label. pixlabel will record the timestamp, row, column, and RGB (red, green, and blue) color values.


Step 5
Download the comma-separated values (CSV) file. You can open this file with a text editor or MS Excel.


Additional notes

  1. If you just want to count objects you can ignore the RGB data.
  2. Image names should not contain commas.
  3. As long as you don't refresh the page, pixlabel will store the records for multiple images.
  4. Make sure you don't misspell the label. If you don't enter any label the app will enter a blank cell in the output file.

If you have suggestions for improvement, want to report a bug, or just want to share how you are using pixlabel, please send me an email to: andrespatrignani@ksu.edu