Introduction to Neural Network Online Project-Based Learning Program
Repo Description:
The latest hurricane - Hurricane Iota, had 61 total fatalities, and 41 are still missing. After a hurricane, damage assessment is vital to the relief helpers and first responders so that resources and help can be planned and allocated appropriately. One way to measure the damage is to detect and quantify the number of damaged buildings, usually done by driving around the affected area and noting down manually. This process can be labor-intensive and time-consuming and not the most efficient method as well. In that case, in the following project, we will find an efficient convolution neural network to identify whether a house was damaged using the satellite imagery as well as the best way to improve it’s appearance.
In this project, I was responsible for:
1. Built a binary-class image recognition ensemble learning model by utilizing mainstream image recognition neural networks, such as Bagging Ensemble Learning Method, Ensemble AlexNet, and GoogleNet.
2. Applied data augmentation to preprocess original data with over 10,000 house pictures.
3. Compared the classification effect of ensemble models, mainstream image recognition neural networks, and each base model in order to highlight the superiority of model results.
Paper related to this project:
[1] Junqiao Fan, Chun Xu, Jiahe Zhang, “An Ensemble Learning Approach of Multi-Model for Classifying House Damage”, in Proceedings of 2021 2nd International Conference on Big Data & Artificial Intelligence & Software Engineering (ICBASE). IEEE, 2021
Note: Authors in this paper contribute equally.
