gale31 / ObjectDetector

ObjectDetector uses OpenCV Haar cascade classifiers to detect different objects in images and videos with Python. For now, this repository includes my trained haar cascade classifier for detecting cars, the default haar cascade classifier for human faces (haarcascade_frontalface_default), a classifier for bananas from CodingRobin and a classifier for wallclocks which are used and tested in programs, detecting the objects from image/video, comparison between different human body parts classifiers and some other programs, which (will) help training the classifiers (for example, a program downloading the "cat box" synset images from ImageNet).

ObjectDetector

ObjectDetector uses OpenCV Haar cascade classifiers to detect different objects in images and videos with Python.

Firstly, I wrote two programs, which used the default haar cascade classifier for faces to find those in images and video. They are located in the return-face folder. Then I used a classifier for bananas from coding-robin and did the same, which can be found in return-banana. After that I repeated the process with a classifier for wallclocks from Celal Caoyn Elgun. The result can be found in, you guessed it, 'return-wallclock'.

Above you can also see results from a 24 second segment of the youtube video "How to Keep Bananas Fresh for Longer". The detected bananas were 3534, and about 35% of these were fake positives.

Then, I trained my own classifier for cars using the UIUC Image Database for car detection. Find more in the folder 'return-car'.

Finally, I trained a classifier for cats in boxes using the ImageNet database (yes, they do have this synset). Find more in the 'return-catbox' folder.

This project was inteded simply as an exepriment and for fun.