Releases: supervisely-ecosystem/model-benchmark
Releases · supervisely-ecosystem/model-benchmark
Internal Release Temp downgrade instance version
v0.2.7 Update config.json
Upgrade docker image tag to 1.0.22 (SDK to 6.73.292)
Custom confidence threshold in supervisely/supervisely#1253
- New feature: Add support for benchmark with custom confidence threshold supervisely/supervisely#1261
- Bug fixes in charts and texts
- Removed duplicated MetricProvider
Upgrade docker image tag to 1.0.21 (SDK to 6.73.292)
- update SDK to 6.73.292 for backward compatibility of the
TeamFilesSelector
widget
Upgrade docker image tag to 1.0.20 (SDK to 6.73.287)
- Bug fixes & improvements (supervisely/supervisely#1249 by @max-unfinity)
- Disable IoU averaging for per-class iou threshold mode
Full Changelog: v0.2.3...v0.2.4
SDK v6.73.284: Add support for IoU per class in eval params; Refactor widgets logic to improve selectors, class matching and task type handling.
- SDK v6.73.284: Add support for IoU per class in eval params; Refactor widgets logic to improve selectors, class matching and task type handling. by @almazgimaev in #16
Full Changelog: v0.2.1...v0.2.3
Upgrade SDK to 6.73.269. Improved evaluation performance for semantic segmentation tasks
Upgrade SDK to 6.73.255. Support comparison for semantic segmentation task type
v0.2.0 Upgrade SDK to 6.73.255. Support comparison for semantic segmentation…
Add option to run speedtest during evaluation
Full Changelog: v0.1.18...v0.1.19
Upgrade SDK version to 6.73.239. Support for Semantic Segmentation. Bug fixes
v0.1.18 Upgrade SDK version to 6.73.239. Support for Semantic Segmentation. B…
Update workflow inputs for model comparison
v0.1.17 Update workflow inputs for model comparison (#13)