Releases: timmh/distance-estimation
Releases · timmh/distance-estimation
Release v1.3.3
implement option to visualize detection IDs
Release v1.3.2
This release fixes a divide by zero error that might occur if calibrate_metric is set
Release v1.3.1
This release changes the sampling process when using SAM to sample the depth at maximum distance from the mask boundaries, which eliminates the percentile parameter.
Release v1.3.0
This release adds the option of using Segment Anything Model (SAM) to provide a finer-grained animal localization than MegaDetector alone.
Release v1.1.3
enable CoreML Onnxruntime execution provider for macOS
Release v1.1.2
This release improves the packaging of the macOS binary, which should reduce memory consumption and startup time. Furthermore, PyInstaller is updated to version 5.6.2 to fix scipy packaging issues on Windows.
Release v1.1.1
ensure all images in a single transect are cropped and resized to the… … same resolution
Release v1.1.0
This release:
- fixes some visual issues with the user interface
- lowers the minimal confidence threshold of MegaDetector v5.0 to 0.05
- uses precomputed depth images if available. This facilitates operation with depth sensors such as stereo cameras
- exports an additional result file which is directly compatible with Distance for Windows
Release v1.0.0
implement github actions build workflow for linux, macos and windows