IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
aa32ff1df34ecb6528dbd107c03e1ba96c3bed19
openvino/tools/pot/tests/NightlyValidationScope.md
Alexander Kozlov bbeec714aa Moved Post-training Optimization Tool to open-source (#7940) 
* Moved POT to opensource

* Added OMZ as a submodule

* Exclude OMZ from ShellCheck
2021-10-15 16:35:35 +03:00

954 B
Raw Blame History

POT tests references for test_nightly.py

All reference numbers for floating-point models were collected on the machine with i9-10920X CPU via Accuracy Checker. Openvino package build that was used - 2021.4.547

  • mobilenet-ssd:
    • map - 0.7251
  • ssd_resnet50_512:
    • map - 0.7966
  • inceptionv1:
    • accuracy@top1 - 0.6981
    • accuracy@top5 - 0.8961
  • squeezenet1_1:
    • accuracy@top1 - 0.5818
    • accuracy@top5 - 0.8062
  • mobilenet-v1-1.0-224-tf:
    • accuracy@top1 - 0.7111
    • accuracy@top5 - 0.8988
  • mobilenet-v2-pytorch:
    • accuracy@top1 - 0.7188
    • accuracy@top5 - 0.903
  • resnet-50-pytorch:
    • accuracy@top1 - 0.7613
    • accuracy@top5 - 0.9286
  • googlenet-v3-pytorch:
    • accuracy@top1 - 0.7769
    • accuracy@top5 - 0.937
  • densenet-121:
    • accuracy@top1 - 0.7442
    • accuracy@top5 - 0.9214
  • octave-resnet-26-0.25:
    • accuracy@top1 - 0.761
    • accuracy@top5 - 0.9256
  • ssd_mobilenet_v1_coco:
    • coco_precision - 0.2332