Data

Tools

Indexes

Applications

Experiments

Open Source Science

https://github.com/azavea/raster-vision-tile2vec

Raster Vision plugin for tile2vec (https://arxiv.org/abs/1805.02855)

https://github.com/azavea/raster-vision-tile2vec

Science Score: 23.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ○
    CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ○
    .zenodo.json file
  • ○
    DOI references
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Committers with academic emails
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (9.8%) to scientific vocabulary
Last synced: 11 months ago · JSON representation

Repository

Raster Vision plugin for tile2vec (https://arxiv.org/abs/1805.02855)

Basic Info
  • Host: GitHub
  • Owner: azavea
  • License: other
  • Language: Python
  • Default Branch: master
  • Homepage:
  • Size: 378 KB
Statistics
  • Stars: 6
  • Watchers: 4
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created over 7 years ago · Last pushed over 7 years ago
Metadata Files
Readme License

README.md

tile2vec Raster Vision Plugin

This repo contains a plugin for Raster Vision that allows for usage of the tile2vec algorithm.

This codebase takes it's tile2vec PyTorch implementation from https://github.com/ermongroup/tile2vec

Codebase Layout

tile2vec

The original code from https://github.com/ermongroup/tile2vec lives in the base package of tile2vec. The code has been modified slightly to work with the Raster Vision plugin better.

One large change is that instead of baking triplets from imagery as a step, and then reading from those triplets at train time, the TileTripletDataset reads from images in S3 directly. The TileTripletDataset takes a set of Raster Vision scenes, and for each triplet, randomly selects 2 scenes - one produces the anchor and neighbor of the triplet, and another randomly selects a distant tile. This procedure assumes that two randomly selected windows of two randomly selected scenes will fit the requirements of a 'distant' tile in the triplet.

tile2vec.plugin

This subpackage contains the Raster Vision plugin, with the following components:

  • Tile2VecTask: Defines a Task for tile2vec. It skips all the chipping steps, since the chips are generated from imagery at train time as described above.
  • Tile2VecBackend: This implants the training code from https://github.com/ermongroup/tile2vec into the training method of Raster Vision, as well as implements prediction that produces embeddings.
  • TileEmbeddings: This is a Raster Vision Labels implementation that maps chip windows to numpy arrays that are the vector space embeddings produced by the Tile2Vec model.
  • TileEmbeddingsStore: The LabelStore implementation for saving off tile embeddings.
  • RemoteRasterSource: This is a custom raster source that allows for completely remote reading of GeoTIFFs off of S3 - this functionality will be merged into Raster Vision as an option, so this should be unecessary with a later Raster Vision release.
  • Tile2VecEvaluator: This is a custom Evaluator that hijacks the evaluation step, usually used to compare predictions to ground truth, to build up a faiss index of embeddings across scenes and chips, along with a csv that can be used as a lookup table of scenes and windows of the index.

Jupyter Notebooks

  • Display Triplets allows you to pull triplets out of the TileTripletDataset and visualize them.
  • Search via embeddings allows you to use the faiss index and lookup table to do a visual search on imagery.

Owner

  • Name: Azavea
  • Login: azavea
  • Kind: organization
  • Location: Philadelphia, PA

Geospatial software engineering for good

GitHub Events

Total
Last Year

Committers

Last synced: about 1 year ago

All Time
  • Total Commits: 1
  • Total Committers: 1
  • Avg Commits per committer: 1.0
  • Development Distribution Score (DDS): 0.0
Past Year
  • Commits: 0
  • Committers: 0
  • Avg Commits per committer: 0.0
  • Development Distribution Score (DDS): 0.0
Top Committers
Name Email Commits
lossyrob r****e@g****m 1

Issues and Pull Requests

Last synced: about 1 year ago

All Time
  • Total issues: 0
  • Total pull requests: 0
  • Average time to close issues: N/A
  • Average time to close pull requests: N/A
  • Total issue authors: 0
  • Total pull request authors: 0
  • Average comments per issue: 0
  • Average comments per pull request: 0
  • Merged pull requests: 0
  • Bot issues: 0
  • Bot pull requests: 0
Past Year
  • Issues: 0
  • Pull requests: 0
  • Average time to close issues: N/A
  • Average time to close pull requests: N/A
  • Issue authors: 0
  • Pull request authors: 0
  • Average comments per issue: 0
  • Average comments per pull request: 0
  • Merged pull requests: 0
  • Bot issues: 0
  • Bot pull requests: 0
View more stats
Top Authors
Issue Authors
Pull Request Authors
Top Labels
Issue Labels
Pull Request Labels

Dependencies

Dockerfile docker
  • nvidia/cuda 9.2-cudnn7-devel-ubuntu16.04 build