Release notes for September 2020
Hey everyone! Here are the details of the major work completed over the last month.
Wildbook DOI for September release!
Wildbook can now be easily referenced for publication. A new DOI will be put out with each month’s release.
- Wildbook DOI: 10.5281/zenodo.4062368
- Related WBIA DOI: DOI: 10.5281/zenodo.4062385
For all Wildbooks
- Multi-feature matching allows for different parts of an animal within an encounter to be matched against, creating a more hollistic understanding of the animal.
- Added multi-process and multi-lane background job compute optimizations.
- Bulk import supports MarkedIndividual.nameX.
- Org admins can run detection and identification for bulk imports for users in their organization.
Silo Security improvements
- Org Admins are now able to delete users from their organization.
- User Management grid now displays Organizations instead of affiliations
Flukebook
- Imported new Right Whale catalog for aerial matching.
- Implemented PIE for Right Whale matching for heads.
- Multi-feature matching set up to support Head, Body, and Peduncle matching of Right Whales.
SeadragonSearch
Successfully rebranded the platform to SeadragonSearch from DragonSearch.
Developers
Cleaned up dev tools that no longer pertain to the platform.
Re-branded Wild Me’s IBEIS project fork as Wildbook IA (WBIA).
Wildbook IA version 3.3.1 released on GitHub (git clone https://github.com/WildbookOrg/wildbook-ia), Docker Hub (docker run wildme/wbia:latest) and on PyPI (pip install wildboook-ia).
Bug fixes
WB-745: Cloning an encounter (such as when multiple annotations are detected in an image) no longer loses the sighting association.
WB-735: Successful detection no longer causes an exception to be thrown.
WB-734: “Match results” option in encounter gallery menu only becomes available after successful detection.
WB-727: Merging individuals handles special characters.
WB-726: Annotations no longer appear to be shifted in the encounter gallery.
WB-696: Merge individuals page handles species correctly.
WB-664: Bulk import no longer creates empty individuals prior to data commit by user.
WB-600: SeadragonSearch email links fixed.
WB-558: Leopard Shark email copy fixed.
WB-517: Bulk import successfully assicates multiple encounters with the same sighting.
WB-442: Python dependency Lasagne updated and incorporated into wbia-cnn, resolving import errors.
WB-304: WBIA no longer retains memory from adjacent background jobs.
B-305: WBIA background threaded jobs no longer locking the Python GIL when in ctypes routines, causing resource timeouts with the front-end web service (moved to multiprocessing).
WB-778: Upgrade to CUDA 10.1 from 9.2 in WBIA Docker images to support GPU-based training and compatibility on Tensorflow for PIE.
WB-121: Fixed WBIA annotation interface broken redirects when using SSL.
WB-825: Resolved build errors in wbia_pyflann and wbia_pydarknet when linking against CUDA 10.X.
WB-846: Revert configuration change to turn HotSpotter super-cache to off by default, fixing ongoing matching issue with ACW and SeadragonSearch.
WB-827: WBIA background job engine re-indexing no longer skips uncompleted and cached jobs.
WB-822: WBIA part annotation interface now stores names correctly on save.
WB-118: Revert time estimate wbia_utool.progress() indicator and move to tqdm for future development.
WB-777: Deepsense orientation predictions no longer fail with updated WBIA Docker manager on Flukebook’s background cache re-compute.
WB-869: Add redundancy and error resolution to INSERT SQL logic with WBIA’s dependency cache with multiprocessing compute and conflicting transactions.
WB-776: MantaMatcher no longer fails to queue PIE background ID jobs.
Coming Soon
Matching for orca, fin whale, and humpback dorsals.
Snow leopard detection and identification.
Projects (BETA), are a tool to allow for catalog distinctions within the platform.