Skip to content

Release v0.3.0

Compare
Choose a tag to compare
@chuneuny-emily chuneuny-emily released this 15 Nov 09:09
· 1031 commits to develop since this release
8fc494e

Added

  • Model Preparation Algorithm (MPA)
    • Add new tasks / model templates for Class-Incremental Learning
      • Instance Segmentation (#1142)
      • Classification
      • SSD and YOLOX model template for Detection (#1156)
    • Saliency map support
    • NNCF (#1157) support
    • HPO (#1168) support
    • Balanced Sampler support for Classification (#1139)
    • Add Adaptive Training for Detection / Instance Segmentation (#1190)
  • Anomaly
    • Add real-life training tests (#898)
    • Add additional check for early stopping parameter (#1110)
    • Add DRAEM task implementation (#1203)

Changed

  • Model Preparation Algorithm (MPA)

    • Replace Class-Incremental Learning models as OTE default models (#1150)
    • Replace OTE ignored label support with external ignored label
    • Enable mixed precision for Classification / Detection / Segmentation (#1198)
    • Enhance training schedule for Classification (#1212)
    • Change Model optimization hyper-parameters for Classification / Detection (#1170)
    • Disable Obsolete test cases for OTE CI (#1220)
  • Anomaly

    • Extend conftest configuration for anomaly backend (#1097)
    • Expose more params to the UI (#1085)
    • Change directory structure for anomaly templates (#1105)
    • Use is_anomalous attribute instead of string matching (#1120)
    • Set nncf version (#1124)
    • Move to learning parameters (#1152)
    • Change OpenVINO MO Command (#1221)

Fixed

  • Model Preparation Algorithm (MPA)

    • Fix inference issues for Detection (#1167)
    • Fix model compatibility issue between SC1.1 and 1.2 in Segmentation (#1264)
    • Some minor issues
  • Anomaly

    • Fix non deterministic + sample.py (#1118)
    • Fix exportable code for anomaly tasks (#1113)
    • Fix local anomaly segmentation performance bug (#1219)
    • Fix progress bar (#1223)
    • Fix inference when model backbone changes (#1242)