As young adults venture into their college careers and engage in partying rituals, they often make mistake of putting incriminating pictures on the internet. Thus, we will use the art of computer vision to minize their problems. We will identify, detect and count beer bottles and beer cups. Next, we will simply draw over it with a picture of an adorable cat or some other unsuspcious items.

Although this is comical and simple problem. The underlying task is quite challenging. This program attempts to detect beer bottles and cups in real-life and practical scenarios (which is very tricky). We have two different classifiers, one for bottles and one for cups. It draws bounding boxes around beer bottles. I train the classifier with HOG features (Histogram of Oriented Gradients) and do left/right FLIPS (for more robust features) and train it on a linear support vector machines (other kernels are too time consuming for this challenge).

Built With

Share this project: