Phogotraphy

This morning I stumbled on to the wonderfully engineered Word.Camera website via a PetaPixel blog post. The premise is simple: Convert a jpeg into somewhat meaningful English language.

To test it out I uploaded this small jpeg of my son on top of an abandoned building, silhouetted by the Sun.

It took the algorithm a few minutes to process an answer. Finally I was presented with several paragraphs of text giving a better than vague description of the image it received:

Of course, a barbed wire, a men, and an energy. Thus, the barbed wire remains unknown. The men evokes typing, and the energy is made from an enterprising or ambitious drive. Probably, the barbed wire remains unknown.

…Yet, a silhouette and a sunset: the silhouette evokes outlining, and the sunset is not the time in the evening at which the sun beginning to fall below the horizon.

Ok, it’s a little jumbled, but understandable English for anyone who has at least a slim grasp on the language. The prose has a familiar air of poetry about it and perhaps with a little human refinement could even be passed off as professional.

I spoke to David Phillips, who operated a poem a day blog in 2014 to ask his thoughts on how the computer algorithm could shake up the industry.

“Word.Camera is an interesting mechanism, which, while some of it’s text is profound, is chiefly without cohesion. As a poet, I am not worrying about my job… yet.”

I tried a few more completely random images to see how word.camera fared, starting with this yellow camera and ending with another little photograph of my son (which to my amusement the algorithm thought was a girl.) Check out excerpts from the text generated below each photograph or click on the photo to read in full.

Here, a render, a sign, and a nobody.

Here, a square, which is made from an any artifact having a shape similar to a plane geometric figure with four equal sides and four right angles. Soon, it evokes paning. In short, it is also known as a tetragon. Indeed, it is an equilateral quadrilateral. In sum, it is not a round. To summarize, it evokes keeling.

…

Finally, a one, a group, and a performance. Thus, the one remains a lonely number.

However, a night, which evokes sleeping time. By all means it evokes sunrising. Notwithstanding, it is also known as a gloom. Otherwise it evokes spending. Indeed, it is the time after sunset and before sunrise while it being dark outside.

…

Immediately, a fly, a north america, and an outdoors.

Too, a gull, a flight, and an egret. Indeed, the gull evokes tricking. The flight may be arrive late, and the egret is an any of various usually white herons having long plumes during breeding season.

…

Too, a football, a protest, and a symbol. Although the football has an offside rule.

Never, a sign, a flagpole, and a disaster.

Then, a nobody, which is not a somebody. Always, it is a single. Again, it is not an everybody.

But a national flag and a flag.

…

Afterwards, a summer, an outdoors, and a landscape.

Nearly, a weather, a light, and a nature. To summarize, the weather may change quickly.

Therefore, a cloud, which has a moisture.

…

In the same way, a child, a girl, and a city. To be sure, the child yearns for playing.

The girl is a young woman, and the city evokes larging.

And yet, the child is not an adult. Finally, it may wet themselves.

In other words, it may smell smoke. Of course, it is prone to jump in excitement.

…

Beautiful, confused or terrifying example of AI? You decide!

---

h/t PetaPixel