EXPOSURE FUSION NEURAL NETWORK

EFNN is a new dynamic range increase solution that aims to enhance image quality in order to achieve close to real time computer vision in extreme lighting conditions.

This new HDR ML based solution attempts to create HDR images at higher speeds than traditional methods.

Get the code & model on GITHUB

Test it now on colab

Checkout the paper

Email the authors

Results :

HDRMark (4286x2848)

Input :

Under Exposed

Normal Exposure

Over Exposed

Output :

EFNN

Mertens EF

AirBellowsGap (4286x2411)

Input :

Under Exposed

Normal Exposure

Over Exposed

Output :

EFNN

Mertens EF

OSTROW (870x580)

Input :

Under Exposed

Normal Exposure

Over Exposed

Output :

EFNN

Mertens EF

SAMPLE001 (1000x750)

Input :

Under Exposed

Normal Exposure

Over Exposed

Output :

EFNN

Mertens EF