Using on pre-trained (on imagenet data) Convolutional Neural Network for object recognition
There are 2 parts to this project. Kindly follow these steps to run the code.
This part does a comparItive study on the different classifiers and gives the testing accuracy. This is only for a single object detection.
Code file - object_recognition.m
This part is the end to end demo of the ORBID system. Given an image, it goes through the following steps -
- Identify the various prominent objects in the given image
- Recognize the objects
- Identify the position of the objects relative to each other (eg., Object A is to the left of ObjectB)
- Build the caption using the object names and position
Code file - load_train.m, ORBID.m
Input image - Input4image.jpg