house-condition-evaluation
source code(Especially image segmentation and network architecture part) reference: https://coursesite.lehigh.edu/pluginfile.php/2315021/mod_resource/content/1/Visual%20Estimation%20of%20Building%20Condition%20with%20Patch-level%20ConvNets.pdf
dataset: image downloads by google street view API address provided by Allentown government
Method:
image segmentation(SIFT, cluster algorithms)
prediction(ResNet50)
Project Member: Kexin, Nathan, Yue
sitf_kmeans_2.py: image segmentation which uses sitf to extract features and using kmeans to cluster. What's more, we use a second kmeans for the keypoints in each cluster which get from the former steps. In this step, we just cluster each keypoint by its coordinate in this step.(Only Physical Distance)