Link to source code

Team Members:

Name: Avni Mehta, Class Id: 11

Name: (Trinadha) Raji Muppala, Class Id: 14

Task Objective:

Perform fuzzy classification and object detection using Random Forest and Decision Tree algorithm. Report on the confusion matrix and accuracy.

Input:

The input for this task is a 10-second McDonald commercial. Three images (front facing man, side facing man and mac donald logo) are taken as reference objects.

Video

Code (Approach)

For the given video and three reference objects, the video is passed through Xuggle API for object detection for the Key Frames and the training features are extracted using SIFT. Next, we run feature extraction for every test key frame where its features are extracted. Combine the features for all test key frames into one file.

The feature vector of both training and testing is passed to the image classification using a classifier (random forest and decision tree). For each testing image, the classifier predicts the weightage from all three labels.

In the end, confusion matrix and accuracy is printed.

Output

The output for Random Forest model.

The output for Decision Tree model.

Observation:

The accuracy for both the classifier is low because of low data, however the prediction rate of Random forest is much better than Decision Tree model.

Contritbution of Team Members:

100% for both since both team members completed the assignment individually.