Good morning. I have a project where the robot recognizes the doors of house. I have succeeded in recognizing them with haar cascade method, but after that, I have become stuck in recognizing the condition of them such as if it is open or closed. Can you give me some idea? Thanks in advance
a binary classifier (such as theCascadeClassifier) can only detect presence or absence of 1 thing.
unfortunately, you have 3 classes already here, door-open, door-closed and not-a-door.
a single binary classifier won't be able to solve this.
probably the best idea is a 2 stage approach, first train a cascade classifier to detect doors in general (both open/closed), against all other parts of a house, then,
train another classifier (maybe an SVM) to distinguish between open or closed door from the detected region .
Oh if I might do a suggestion, if your dataset of open <--> closed doors is limited and SVM does not reach satisfactory results,try a simple NaiveBayes classifier. It keeps amazing me how powerful it is in cases with less data. And its quick at evaluation time!
@AlexGr I am also working on a similar task and looking for sample dataset for doors' images. Can you share with me any repository from where I can get lots of training images of doors, or maybe a pre-trained xml file for doors ? Thanks
@muhammad.umar600 Dear Muhammad i send you the link with photos. https://github.com/ESDA-LAB/Door-dete...
Asked: 2017-07-29 05:23:13 -0600
Seen: 1,872 times
Last updated: Jul 29 '17
