Microsoft Edge is enriched with artificial vision for images

Microsoft added on edge
a new feature improving the accessibility of its Navigator
for visually impaired and blind people who use a screen reader. It is now possible to obtain alternative texts automatically for images that do not have any to facilitate the understanding of a web page visited.

By being processed by a special algorithm, a large number of images will now be able to bring more context to web pages for the visually impaired. But Microsoft warns that it’s always better for the authors themselves to fill in the alt text in the images.

An important accessibility option

According to Microsoft, at least half of the images that pass through a screen reader do not contain alt text and therefore cannot be read by the software, causing certain pages to lack context. Image alt text is very important but is often overlooked when an image is uploaded to a page in favor of captions. The caption of an image is its written description but only the alternative text is taken into account by a screen reader.

Thanks to machine learning, Microsoft Edge now makes it possible to obtain alternative texts automatically on images that do not have said texts. If the user gives permission, the image without alt text will be sent to the special Computer Vision API which will analyze its overall content. The algorithm will then output an automatically generated placeholder text.

The technology is not free from flaws and Microsoft warns that the description will not always be of quality but tempers by saying that for a person using a screen reader, “ having any description is better than no description at all ”, before adding that this functionality is an alternative but that it is always preferable that the alternative texts are filled in by the authors themselves for more precision.

A feature at the start of deployment and therefore limited

To activate this feature, click on the three small dots at the top right of the Edge browser, go to Settingsselect Accessibility then activate “Get image descriptions from Microsoft for screen readers”. Whenever an alt text has been auto-filled, the reader will first say “Seems to be” or “Seems to say” if it describes text on the image.

This accessibility feature is currently only available on Windows, Linux and Mac but not on Android and iOS. The API can output a description in five different languages ​​at the moment (English, Spanish, Portuguese, Japanese and Simplified Chinese) and supports the majority of image formats that can be seen on the web.

But a few exceptions come in: images marked as “decorative” will not be processed since they do not provide any particular context, images below a size of 50x50p will not be analyzed either, just like those ” excessively broad” or categorized by the API as gory, pornographic or sexually suggestive.

Source: Microsoft
, Softpedia