Term: Spring 2018
-
Team 10
-
Team members
- Xinlei Cao
- Wenyuan Gu
- Wenshan Wang
- Yiyi Zhang
-
Project summary: In this project, we have created a classification engine for images of poodles, fried chicken and blueberry muffins. Our goal is to improve the prediction accuracy from the baseline model (GBM with decision stumps on 2000 SIFT features) and to enhance computational efficiency in terms of running time, storage and memory cost. We have examined feature descriptors such as cale-invariant feature transform (SIFT), histogram of oriented gradients (HOG), and local binary pattern (LBP), and evaluated classifiers including Gradient Boosting Machine (GBM), RBF Support Vector Machine (SVM), Random Forest (RF), XGBoost, Logistic Regression, Convolutional Neural Network (CNN). Logistic Regression with LBP is finally selected as our improved model, which increases the testing accuracy by 9.5% from 67.3% to 76.8%, and significantly decreases the training time from 403.53s to 0.33s.
Contribution statement:
- Xinlei Cao: Established and optimized the Convolution Neural Network (CNN) in Python. Completed the data augmentation such as flip and rotation, thus improved the generalization ability of the model. Finished the connection of Rmarkdown and Python.
- Wenyuan Gu: Established and tuned Logistic Regression, Random Forest, and SVM model using SIFT, LBP, HOG features, respectively. Created presentation slides and presented in front of class.
- Wenshan Wang: Established and tuned GBM, XGBoost, and SVM model using SIFT, LBP, HOG features, respectively. Compared training time and accuracy of all models using LBP and HOG features. Wrote main.Rmd file.
- Yiyi Zhang: Provided direction on how to proceed the project. Extracted features using SIFT, HOG, and LBP and completed the feature extraction and project summary part of the main.Rmd. Updated README.md.
Following suggestions by RICH FITZJOHN (@richfitz). This folder is orgarnized as follows.
proj/
├── lib/
├── data/
├── doc/
├── figs/
└── output/
Please see each subfolder for a README file.