PyrΔTE

The PyrΔTE interface allows a user to input their own CIELAB color values, either individually or as a series of data points in a .csv file.

 

https://mineralcare-calculator-htm4lvshfq-nw.a.run.app

How to Use

L*a*b* Fields

Input your color data into the corresponding field

Calculate

Click this button to calculate the specimen's T-Score, ΔE*, and pseudo-object color.

Reset

Click this button to clear your input L*a*b* values.

T-Score

This field produces the likelihood of an untrained observer to notice tarnish.

(See the T-Score tab for more info.)

ΔE*

This field produces the ΔE* (color difference) between your specimen and the ideal untarnished specimen.

(See the ΔE* tab for more info.)

Color

This field produces the pseudo-object color of your specimen.

(See the Color tab for more info.)

Upload

Click this button to upload your .csv file of color data for batch processing.

Loading

The 'Mineral Care' icon and text 'Pyrite Decay Calculator' in the upper right-hand corner of the app will change to a loading wheel and the text 'Loading...' once you have uploaded a .csv file.

Both the wheel and the text will remain until you're file has been processed and your results have been downloaded.

 

  1. Enter your own CIELAB values in either the 'Initial' or 'Current' color fields.
  2. Press the 'Calculate' button
  3. Wait briefly for your results.

 

If you enter CIELAB values into both the 'Initial' and 'Current' color fields, the program WILL NOT produce a color difference value (ΔE*) between the two entered CIELAB points. Rather, it will provide T-Scores and ΔE* values comparing your values to the ideal untarnished pyrite specimen (see ΔE* tab for more details).

This is something I hope to address in a later version of the program.

A .csv file of color values can be uploaded using the upload button (arrow pointing upwards) in top right-hand corner of the app.

It will take some time to process your data, so please be patient and do not refresh or close the page. Your data will be lost if you do so.
Processing time depends on file size. Large files could take up to a few minutes to complete.
E.g., A ~3kb file will take less than 1 minute to process. A ~15kb file will take 2-3 minutes to process.

Once processed, a .csv file of your results will appear in your 'Downloads' folder.

 

NB. Batch processing is limited to specific web browsers. If you upload a file and do not see the Mineral Care icon turn into the loading icon, please try using a different browser.

Suitable Browsers Unsuitable Browsers
Microsoft Edge Mozilla Firefox
Google Chrome  

 

T-Scores indicate whether an untrained observer (such as the average museum visitor) will notice whether a specimen of pyrite is tarnished.

Specimens were visually examined by Kathryn Royce, the curator or collections manager, and the volunteer who collected the color data (if applicable). Each individual recorded whether they perceived the specimen to be tarnished (1) or untarnished (0). These values were then averaged, resulting in the T-Score.

ΔE is the notation used to indicate the total color difference between a pair of color measurements as plotted in a three-dimensional color space (like CIELAB). When using CIELAB, ΔE* values are expressed in CIELAB units.

With PyrΔTE, we are comparing your measured color data to that collected from our ideal untarnished specimen. This applies for both single and batch processing.

Our ideal untarnished specimen is OUMNH-MIN.28008 with CIELAB coordinates 76.26 L*, -0.78 a*, 12.39 b*.
These values were derived from the average of ten measurements.

 

NB. The ΔE* values produced by the app are 'predicted' by the algorithm, rather than being calculated with the CIEDE2000 equations. The predicted values are very close to those produced by the CIEDE2000 equations. 78% of the AI ΔE* values fall within one standard deviation (or ± 0.1 CIELAB units) of the CIEDE2000 value. This, I recommend rounding the decimal to the nearest whole number (e.g., 24.67 => 25).

If you want more accurate values, or to directly compare your before to after values, you may wish to download the excel add-in, ColorTools, created by Edgardo García. This add-in includes the following functions:

  • DeltaE - ΔE calculation using 1976 CIE ΔE formulae.
  • DeltaE94 - ΔE calculation using 1994 CIE ΔE formulae.
  • DeltaE94Textiles - ΔE calculation using 1994 CIE ΔE formulae with suitable parameters for textile industry, according to Bruce Lindbloom.
  • DeltaE2000 - ΔE calculation using 2000 CIE ΔE formulae.
  • DeltaECMC - ΔE calculation using CMC ΔE formulae with parameters l and c.

The colors produced in the 'Color' square are called pseudo-object colors: colors which represent measured color data. Despite being largely ignored in the color science and heritage literature, pseudo-object colors are incredibly useful (especially for the non-expert) in determining the acceptability of the measured numeric values in representing the object’s perceived color.

 

These colors are rendered in sRGB. You can copy the RGB or HSL values for the produced pseudo-object color by clicking on the 'Color' square.

Expand All

PyrΔTE is the result of  a pilot study using pyrite color and tarnish data to determine whether AI can be used to help identify change in museum specimens. Whilst this program is presently limited in scope to colorimetry and pyrite, PyrΔTE demonstrates that, with further development, similar AI tools can be created to aid identifying and treating visual and material changes to museum objects.
 
Data was collected from hundreds of pyrite specimens from Oxford University Natural History Museum, National Museum Cardiff, National Museums Liverpool, and the Sedgwick Museum of Earth Sciences. Nineteen volunteers helped to collect over eleven thousand data points. Both the volunteers and the collections’ curators also assessed whether each specimen was either tarnished or untarnished.
 
The collected data was then handed over the project partner, OR3D, who fed the data into two separate Regression AI modules in Python (TensorFlow and the Keras module) to identify patterns within the dataset. Here, the AI used the color data to calculate tarnish likelihood and the overall color difference (ΔE*).
 

The latest version of the calculator, dubbed PyrΔTE, was completed in June 2022.

Papers

  • work in progress

 

Conference Presentations