Term: Fall 2018
-
Team #
-
Team members
- Peilu Zhang([email protected])
- Chun Zhai([email protected]）
- Qingyang Zhong([email protected])
- Rui Zhang([email protected])
-
Project summary: In this project, we created a classification engine for enhance the resolution of images. We used one GBM baseline model, XGboost and Random Forest model to achieve this goal. And we also use more neighbors to extract the features to improve our baseline model. Finally we choose 8 neighbors feature extracting and XGboost model to predict the high resolution pircture from low resolution pictures.
Contribution statement:
Peilu Zhang: " I write all the baseline code, including feature.R and superresolution.R for all the models and small changes in test.R and train.R. I shortened the feature extracting time of 1500 pictures to 3 mins, tuned the gbm parameter by cross validation. And I fixed the code of xgboost model."
Chun Zhai: “I am responsible for the improved feature part of the project. This includes writing codes for 24 neighbors feature and testing it on both GBM baseline model and XGBoost model, tunning the parameters through cross validation. I also write codes in Python for SIFT key points detection method and try other feature improving methods like padding other constant and adding diagonal neighbours."
Qingyang Zhong:" I'm responsible for improving model using RandomForest algorithm and write all parts in RandomForest model. I tuned parameters by cross validation then tuned the quantity of data to balance speed and MSE. I also write the code for calculating PSNR. And help with preparing for presentation."
Rui Zhang:"I'm responsible for improving the model using XGBoost algorithm and I tuned the parameters by cross validation. I also need to compare all the possible improvement methods on theoretical level. And I summarized the performance measure indicators regarding accuracy, memory and speed."
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.