Almalence Super-Resolution improves image data, allowing the recognition algorithms to work robustly and confidently in harsh conditions.
Machine vision strongly depends on the quality of input image data. Insufficient resolution due to long distance to the object, or high noise due to dim lighting drastically decrease the robustness of recognition, with the neural networks producing unconfident and often wrong results.
Almalence applies Super-Resolution to increase the resolution and reduce the noise, thus providing higher-quality input image data to the machine vision algorithms.
With Super-Resolution, surveillance and security cameras can recognize faces and license plates, automotive cameras can recognize traffic signs and road objects from longer distances and in low light.
As Super-Resolution is applied to the small region of interest only, it can run on virtually any camera, including those with low computing power.
Below is a proof-of-concept demo, showing how Almalence Super-Resolution improves the reliability of face recognition of OpenVINO/Movidius platform with a FLIR security camera:
Right: OpenVINO as is. Low confidence and wrong guess.
See the demo video:
