Hi there,
I am working on
matlab, particularly in object categories recognition.
The work, in other words, is about recognising multiple object classes (object classes recognition), where the dataset that I am using is MSCOCO.
I have done the detection using Faster RCNN as implemented in matlab, and have already got the detection results. Now I am looking for help to implement a context model (a mode that can help detector to rescore the detection results based on the relationships among objects known from the annotation). For example, Image (I) has a scene containing sky, airplane and ground. When the detector, by mistake, detects the airplane as a car, the model should improve the detection results to be an airplane, as the probabilities known from the annotation is that airplane is likely to occur in the sky rather than a car in the sky. Please note that I have already encoded the relations among the objects, and now I am just looking for someone to help me to implement and develop a classifier that can help to improve the detector’s results. The faster this work can be achieved the better it is.
Here are some papers that I am following, which clearly explain the aim of my work
http://people.csail.mit.edu/myungjin/publications/treeContext.pdf
http://people.csail.mit.edu/myungjin/publications/outOfContext.pdf
Thank you very much for your help and time,
I am also looking for someone who can do the above described job, plus be with me when needed, because I have just started this on-going work, and I will need further help