Skip to content

v0.1.0 release

Pre-release
Pre-release
Compare
Choose a tag to compare
@rikenmehta03 rikenmehta03 released this 19 Mar 14:56
· 18 commits to master since this release
v0.1.0
1ef827c

first release of imsearch.

What's included?

  • The framework includes API to create search-index and add images with single and bulk mode. Both local image path and URL are supported.
  • Backend uses Yolo-v3 object detector trained on COCO dataset
  • The framework search uses object wise similarity with cosine distance in resnet50 feature space.
  • An efficient cross-platform similarity search library NMSLIB is used.
  • Redis is used as a messaging queue between feature extractor and core engine.
  • MongoDB is used to store the meta-data of all the indexed images.
  • HD5 file system is used to store the feature vectors extracted from indexed images.

Plans for v0.2.0

  • Add unit-tests using nose2 & unittest testing framework
  • Add support for cloud storage services to store images (AWS S3, Google Cloud Storage)
  • Add support for multi-threading or multi-processing in feature extractor back-end.
  • Add documentation
Assets 2
Loading