Friday, November 14, 2014

Human Posture and Dynamic Scene Analysis

As a street photographer, I am faced with crowded scenes surrounded by busy backgrounds. Posts, cars, buildings, lights are the most common distractions when composing a picture. And since my genre of photography needs people in postures that tell a story, it is easy to dilute, obscure or lose that story in a frame overwhelmed by the environment's distractions.

This challenge is, I believe, greater when B&W film is used since the human eye can detect contrast, differences and depth of field better when looking at a color picture. After all we are trained to look at the world in color!

Yet, the interest of isolating humans through their shapes and gestures in pictures and videos is often encountered in the vast fields of technology and sociology. Indeed, informatics science is ripe with algorithms that “interpret” video captures to identify specific human behaviors in crowds based on their appearance but also postures. Interestingly, geography and culture dictate certain appearances and postures, so algorithms work best when incorporating these local shape cues and postures. Think about surveillance videos and how they are interpreted. But also think about how environments can be design more effectively say in cities, for facilitating circulation, ambulation, and safety. And that is where sociology comes to play a significant role because while there are global postures we all understand, there are a myriad of "local postures" that leave us clueless or worse, predispose to erroneous interpretation of the scene.

… I recall an early morning walk in Taipei when for the first time I saw an older lady clapping her hands while sitting on the ground behind a bush. I thought she was calling for help so I rushed to help her. Later, I was told she was practicing Tai Chi…  It was 1978, in Calcutta, India when I saw many people laying on the ground in the square, looking perfectly dead.  I learned that there were public places to smoke Afion (opium) and practically pass out for hours. A video of that square would have given many a totally macabre impression.

So, I was reading about the use of such posture-clue algorithm to build computer simulation  for driver assistance. One can imagine that this algorithm will change from country to country depending on the road architecture, the way traffic is supposed to flow, the lighting of the streets, if people just cross busy streets by running across, etc.  I was intrigued by the lack of B&W photography discussion although one article(1), by Rodriguez and Shah, presented convincing scientific evidence about the importance of local postures when the human eye tries to identify people in a crowded scene.

That was enough to make me decide on a specific street photography theme at my next outing: could I identify local postures that most of my readers would immediately identify and interpret with little variation?
I took my 1970’s Nikkor 105, 2.5, ASA 100 film and off to town I went!

… I am in a cowboy town where many statues attest to its glorious past and cherished present. The first picture I took was perfectly framed for me: a local man wearing a cowboy hat was about to pass under a cowboy-on-horse statue. I waited for a split second to get the bronze and felt cowboy hats touch just so, and wanted the head of the cowboy and horse of the statue to align perfectly with the cables above. Why? Because this was an “inverse theory” I wanted to test:  when the background were harmonious, would the viewer of this picture notice the cowboy hats first or the horse’s head?


Then I walked around for a while without clicking once. All seemed to be well organized, almost ordinary. I needed an angle, a story that a B&W film can capture.

And I saw this woman through the space the bodies of two men had left open. Her long fingers and impeccable make-up were eye-catching and vibrant, and her cowgirl leader jacket just perfect for that local posturing clue. So I took a couple of shots.




Now I wonder if a visitor of my blog from Asia will see the same thing as a viewer from Europe or the U.S. What would the second picture of the woman mean? Is she scolding the men? Telling them how lovely it is to have sunshine in November? Or would she be seen as just a person in the street unknown to these two men?

Further, would a viewer from a non-cowboy town even notice the cowgirl leather jacket to identify a person and then notice her hands and face in the first picture?

To test, I ran these pictures by a few friends. Most looked at the horse first; and everyone stated that they would have liked to see her full face… I suppose that was a good sign meaning that they did notice her posture hidden behind the out-of- focus silhouettes of the two men.
But they knew I was in a cowboy town and perhaps that they were disposed to noticing the jacket.

No matter what, I was surprised that I was the only one in this small sample of reviewers to have focused on the cowboy hats first….

(1)http://vision.eecs.ucf.edu/papers/ACM-MM-Rodriguez.pdf


November 14, 2014
© Vahé A. Kazandjian, 2014

No comments:

Post a Comment