CS 5542 BigData Lab Report #06
SPARK PROGRAMMING - Machine Learning Tasks : Fuzzy Classification + Object Detection Task (Decision Tree + Random Forest)
Create your own dataset for Image Classification/Object Detection Problem. Handle fuzzy classification/object detection task using at least two classification algorithms. Report accuracy and confusion matrix obtained.
- Object Detection - Generate Key points and features from the video clip. Train each image set by the its individual image.
- Object Feature Extraction - Obtain features for each image category separately.
- After training and testing, we can build models - Decision Trees and Random Forest.
- Report Accuracy and Confusion Matrix for each model.
Chameleon, Fish, Peacock, Polar Bear and Sea. ( Only select some of them show up in the video. )
Process the videos with SIFT and train the data sets and obtain features from the clip. Get the features from the images so we can have some training data. Test out images based on its category and build up models.
The accuracy is low for a couple of reasons: the video is not long enough to gather enough training features. Too much noise in the video ( Lots of objects shifting and changing ). Thus, if we want to improve the accuracy in detecting images in my dataset, I need to find a video that is long enough and pure enough to have a solid feature training.
Write an Android Application: Image classification(fuzzy) & Object detection through a) Spark API. b) Clarifei API
- Create UI Interface for client - buttons, text box for showing the result of certain image, an image box to scan
- Build up app to connect with API for Spark and Clarifei.
Side-note: Not quite taking out all the errors yet.
