There is no denying to our long time fascination with the elite forces Hollywood movies. The whole world seems to be crazy about the abilities of the law enforcement agencies. One of these knacks that grab our attention strongly is how the Special Forces zoom in on grainy photos and security camera footage. It’s because it’s not easy to enhance the pixelated image like we see in movies.
This seems impossible in real life as in small number of pixels there is not enough detail to bring out a clearer image than the original one. But now Google has found a way to make its users act just like some law enforcement agents.
The above set of images explains the pixel story in detail. The images on the left show the 64 square pixel source images. The images in the middle show what Google Brain new system can build for them. The higher resolution images are represented by the rightmost column and users can compare it to the middle column to see how efficient the Goggle system is.
A pixel recursive super resolution model has been created by Rayn Dahl, Muhammad Norouzi and Jonathan Shlens who came up with the idea that can synthesize details in low-resolution photos with the help of combination of power of two neural networks.
The conditioning network initially tries to map the 8×8 source image against similar high resolution images. It even approximates what the actually looks like when it’s zoomed in. After that some realistic details are added to the final output. This is attempted by learning what each pixel in a low resolution sample generally matches to in high resolution images.
Right now, the system seems fairly proficient and come in handy for assisting special services in solving crimes. But as these images are computed on the basis of approximations so they are not considered real therefore they cannot be provided as evidence for any investigation processes. But it’s amazing how technology is bringing science fiction closer to reality. And unlike those science fiction movies where technology resides only in life like android bodies, it may soon neatly fit in our small devices.