Please note that the online version of the algorithm accepts jpeg images only.
Please also note that the online version of the app will only accept images between 40 KB and 1.2 MB.
The image of the painting must be without frame and must be the full image of the canvas.
Procedure if the image has been saved to your computer:
- Click ‘Browse’. This will open a directory on your computer from which you can select an image. The name of the file selected will appear to the right of the ‘Browse’ button.
- Click ‘Upload Image’. This may take several seconds. You will know that the file has been uploaded when the name of the file to the right of the ‘Browse’ button disappears and is replaced by ‘No File Selected’.
- Click ‘Score Image’. The result will be a score that varies around an average of ‘0' (0 is the score of the average painting in the corpus. For a description of what each score level means, go to the ‘About’ page.
Procedure if the image is from a website:
- Copy the URL of the image including the .jpg at the end.
- Paste the URL into the textbox below the ‘Browse’ button.
- Hit select ‘Upload Image’ and wait for the image to upload. This may take a few seconds.
- Click ‘Score Image’. For a description of what the score means, go to the ‘About’ page.
The quality (compression) of the jpeg image is important. A jpeg image that has been resized has lost some information. Scores generated by the algorithm for the same basic image can diverge by as much as 10% from the correct score if a low-quality (resized) image is used. As such, images that are derived directly from the raw camera images are likely to be of better quality and the score generated from them is likely to be more accurate. If you get different scores for different images of the same painting, the higher quality image is likely to be the best indicator of the actual score. Alternatively the average of the different scores can give a good indication of the true score.
The hue of the image is important. Photographic images saved as jpeg images can diverge from the original image in terms of hue. This can affect the score generated by the algorithm by as much as 5%. Therefore, it is important that the hue of the image is as close to the hue of the original painting as possible.
If you get a score that does not seem to fit the painting (for example, if an Old Master’s painting scores below 1.5 or a badly executed painting scores above 2), try a different jpeg image of the same painting. You may find that there is a difference in the score due to problems with the initial image.