Skip to content

Image visual similarity based search in NAIP satellite imagery of some areas of California using Deep Learning. Deep CNN provides output features that are converted into 512-bit binary vector that compactly represents the input image and can be used for similarity based search.

License

Notifications You must be signed in to change notification settings

sdhayalk/Large_Scale_GeoVisual_Search

Repository files navigation

Large Scale GeoVisual Search

Implementing image visual similarity searching and indexing in NAIP GIS satellite imagery of lower California using Deep CNN. Trained ResNet architecture on UC Merced Land Use Dataset with data augmentation to learn geographical imagery features. Truncated last layer to get output features that are converted into 512-bit binary vector that compactly represents the input image from NAIP satellite image after forward pass, This vector is used to calculate hamming distance for similarity based search.

Further aim to parallelize searching across multiple nodes in AWS cluster using Hadoop/Spark.

Project is onging and is not yet completed.

About

Image visual similarity based search in NAIP satellite imagery of some areas of California using Deep Learning. Deep CNN provides output features that are converted into 512-bit binary vector that compactly represents the input image and can be used for similarity based search.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages