The 2018 Conference on Computer Vision and Pattern Recognition (CVPR 2018) was held last week. CVPR is one of the world’s top conferences in machine learning, especially in the field of computer vision. It was highly anticipated, and some interesting research was presented there.

One paper I particularly liked is Learning to See in the Dark. The researchers created a dataset of images taken in the dark and with normal light settings and trained a Convolutional Neural Network to perform image improvement for the picture taken in the low-light settings. The results are astonishing! Video is worth thousands of words, check it out:

The authors compare their method to existing denoising, deblurring, and enhancement techniques. In my opinion, it almost always performs much better. And it takes about a second to process! I definitely see it implemented in the smartphones in the nearest future, since every decent device can run neural networks onboard.

More and more people are studying machine learning and neural networks now, and we see plenty of applications proposed which in turn motivate more and more people to learn machine learning and apply it to different tasks.