We used to see major improvements in smartphone cameras year after year, but this seems to have changed recently. Manufacturers are now struggling to create and integrate cameras on smartphones that produce higher quality photos from last year’s model. The main reason for this deceleration in the smartphone camera evolution, is due to hardware limitations. Built-in cameras need to be small and compact, leaving them lacking in comparison to DSLRs with large and interchangeable lenses. For this reason, smartphone cameras will never be able to catch up with the quality of DSLRs. Or will they?

Smartphone manufacturers are well aware of these hardware limitations. Companies like Google have turned their focus on improving their camera quality through the use of software. The results of their efforts are astonishing and can be seen on their latest phone, the Pixel 2. Despite being the best smartphone camera in the market so far, Pixel 2 falls behind even the lowest quality DSLRs.

Does this mean there’s no hope for smartphone cameras afterall? Well..no.

A research team has created a tool that helps you capture photos with quality comparable to that of a DSLR. They achieved this by using deep convolutional neural networks to train a model that is able to recreate missing details on photos taken from smartphones. This neural network is trained on a dataset of images captured by state of the art DSLRs as well as smartphones. This way, it is able to tell what pieces of information are missing from the photos taken on smartphones and therefore recreated those details and add them to the low quality photos. The final images produced by the algorithms have quality comparable to that of an original photo taken from a DSLR camera.

You can test this yourself by uploading your own photo on their website. If you’re interested in finding out more about this research, you can do so by reading their paper, which is available for free on their website.

The post Enhance the quality of your photos using Artificial Intelligence appeared first on BadFive.