Skip to content

Releases: microsoft/aerial_wildlife_detection

AIDE v2.2

19 Nov 10:49
Compare
Choose a tag to compare

AIDE v2.2

Includes the following fixes:

  • Fixed a bug in the Debian/Ubuntu installation script related to the AIDE worker daemon process
  • Adjusted the image viewer for better scrolling, zooming, and panning
  • Fixed password update check during AIDE startup

AIDE v2.1

04 Aug 13:23
94f06db
Compare
Choose a tag to compare

AIDE v2.1

  • v2.1 introduces a new installation script for Debian/Ubuntu systems. This has been tested on Ubuntu 20.04 LTS but is still in beta mode. See here.
  • Many bug fixes for dependency versions, migration script, UI (new project screen), model import in Model Marketplace

AIDE v2.0

25 May 07:19
Compare
Choose a tag to compare

Official release of AIDE version 2.0.

This release contains the following improvements over the previous v1.9:

  • New models based on the Detectron2 framework:
    • Image classification: AlexNet, DenseNet, MNASNet, MobileNet, ResNet, ResNeXt, ShuffleNet, SqueezeNet, VGG, Wide ResNet
    • Object detection: Faster R-CNN, RetinaNet, TridentNet
    • Semantic segmentation: DeepLabV3+
  • Model Marketplace:
    • Share your models across projects with two clicks
    • Download trained models to disk, upload to AIDE or import from the Web
    • Manual label class assignment for pre-trained models for maximum performance (in addition to dynamic model adaptation)
  • Preparation towards code framework for AIDE model zoo (stay tuned for release!)
  • Many bug fixes regarding the Workflow Designer, data management, and more

AIDE release for scientific paper

11 Sep 11:27
Compare
Choose a tag to compare
Pre-release

Pre-release of AIDE 2.0 with all functionalities from the paper (Kellenberger, B., Morris, D., Tuia, D.: AIDE: Accelerating Image-Based Ecological Surveys with Interactive Machine Learning. Methods in Ecology and Evolution, in press).
Functionalities contained:

  • Labeling interface (image labels, points, bounding boxes, segmentation masks)
  • Built-in models (ResNet, RetinaNet, U-Net) with active learning support
  • Workflow designer
  • Data management
  • User and model performance evaluation

Not yet built-in, resp. work in progress:

  • Model marketplace
  • Improvements to existing functionality
  • New, unannounced capabilities