Experimenting with OpenCV & Selenium
At first, I tried the idea of surfing web and using selenium to detect real time information of the web to visualize in ASCII to sort of replicate machine’s vision of data, which felt sort of meaningless at first, however I was thinking how I could expand the idea of extracting or identifying images through the net or perhaps detect fake images?
However, I did try to experiment with just creating an ASCII overlay on top of images only, but it took too much process to load and ended up crashing the web, due to the amount of data upon loading the page. Well, other than just aesthetic purposes, I felt that I had to move on from here.
Heatsensor Tracking
Hence, moving on to the next trial, I tried the function of heatsensor with OpenCV where it detects the eye and movement which sort of record them in a separate window. The initial intention was that to display the trails of data recorded and in your face manner.
But offcourse, it lacks the impact and originality since it's just data visualization which relates to nothing. Therefore, having remembered how Ashley shared her project experience in integrating values that would matter to us humans.
I hence tried to ask ChatGPT to simulate a calculation to how much energy it will use to record all the data in data centers with the heatsensor detection. Very quickly, I realised how it was irrelevant to my topic of data privacy as this data log integration is more related towards a sustainable future with tech. Eventhough it was irrelevant, I thought I would use the data log for something else to notify people with other informations that could relate to data privacy.
Thoughts
From this exercise, it does not really relate to my topic, however there are some takeaways from the experiment itself by presenting the data logging to provoke the sense of intrusion to personal space.
New Experiment with Yolov11
Moving foward with a new experiment, using the same application but with the addition of Yolov11, the integration of Yolov11 was to use their detection capibility as well as library for more accurate detection. Therefore with existing library for the detection, I don't have to train a model from scratch and I could train my own model if I wanted to.
The inital set-up started from capturing live feed from the webcam and leaving trails of image captured in random noise steps. Subsequently with the addition of the data counter on the top left corner. Followed by having the Yolov11 integrated into the code to identify object that comes within the webcam. However, in the process the random noise appeared as an overlay glitch effect which was not I really wanted and I realised that having all effects happening in one window may not convey the message of real-time data. Therefore, I added an additional window with one that shows the webcam live feed
In the process of adding the new code for the additional window, I was thinking of how I could use Yolov detection function and create an mosaic filter effect upon detection, reason being to create that reverse diffusion concept of diffusing the image data before reconstruction.
However, I wanted to refine and polish the reconstruction method using python. I did not want to use any kind of AI model for it and hence using my own interpretation, I had ChatGPT to help me reconstruct the code in such that the pixel that relevance of the color beside it to reconstruct the image data taken. The reason for not using AI models was that it would have make no difference from just a real-time generative art which I did not want to create, I wanted to show the "truth" or what's behind these AI generative art.
As I was adding the new code, I wanted the reconstruction visual process to be less random, hence the grid base animation helped to give that growing effect. On top of that, I changed the size of region captured for the reconstruction window as I thought that it will be more interesting if people take a closer look, they realise that the pixels are data captured from real time.
I wanted to push the visuals and tried adding RGB pixels to give the processing effect but the result was not really great as it takes away too much attention and looks chaotic.
