After my in-depth interview I needed to test my sketch with a wider audience. I play roller derby and I decided to take my laptop with me to a free skate session to test my sketch out on my team mates. As most people were busy with skating I did not want to disturb them unnecessarily, and was unable to conduct any in-depth interviews as I would have liked. However this did give me a chance to test out a sketch with a group of people who were moving at a very fast pace. This might be important as during my observations of the foyer space, it was a very fast moving environment. If my piece could stand up to the rigors of a fast moving pack of skaters then hopefully the foyer will be handled easily. In addition the use of rigorous interview testing was somewhat contrary to my stated aims of the piece, as I stated that I did not mind if people did not pay full attention to it.
One of the main problems I had with setting up the sketch was finding a suitable place to set my laptop. There were chairs or tables in the sports hall and so I had to balance my laptop on a bag. This made it unstable but also meant that the camera was not in the optimal position to pick people out. During testing I also had to make sure that some people were not filmed, as they did not want to be “close-up” on camera. Thankfully the lens of the camera on my laptop is quite low resolution so it could not pick out their face. I also made sure not to point my phone camera into the sports hall.
One of my main concerns before testing was if the facial recognition would work with people wearing helmets. To my surprise it worked fine. However there were two cases were the facial recognition failed which I did not anticipate. Some of the men I tested had beards, which meant that the facial recognition did not pick them up. I also tested Emma, who is in the video and the facial recognition seemed to be intermittent picking up her face as well. I suspect this is because she was wearing glasses. Unfortunately there is not much I can do to remedy this. The fault is within the algorithms of face detection in Open CV itself. A face is identified by the ratios between certain features (such as the eyes nose and mouth) (Open CV, 2014). If something were to obstruct this, such as hair or glasses the face will not be detected. Conversely if an object or structure has differences in colour in the same positions as these ratios, it will be detected as a face.
A face is detected on the skaters in the distance and the birds are flocking to them.
On the other hand, I was surprised at how well the faces were detected faces that were in fast motion and were quite far away. The piece also impressively handled well the quick direction and speed changes of the skater, and the bird flock followed them accordingly. This is also a good sign for the future of my project. Unfortunately I didn’t get to ask too many people on the thoughts on the project (they were too busy skating!) but on the whole the response I received was positive.
With these tests completed I feel that I am ready to put the piece into the foyer.
Open CV, 2014. Face Recognizer with Open CV. [online] Available from: http://docs.opencv.org/modules/contrib/doc/facerec/facerec_tutorial.html [Accessed 17.01.2015]
Thank you to Emma “Leaf Fairy” Pickworth, and DKRD for participating in the testing.