[ENHANCE] Parametrize saliency maps dumping in export parametrizable

[GalyaZalesskaya]

Added dump_features parameter in otx export not to calculate and dump saliency maps and representation vectors for each export.

Checked export and inference of pth and IR model for:

• Classification task (classic classification - MobileNet-V3-large-1x)
• Detection task (ATSS model)

Expected result: improved efficiency of deployed models if no explanation is needed

TODO

1. Implement dumping feature according to flag for Segmentation task, Action Detection and Anomaly task
2. Tests

[negvet]

Minor comment from my side

[negvet]

Unit tests are required, although can be added in a separate PR

[JihwanEom]

Could you also check pre-commit issues?

[GalyaZalesskaya]

@JihwanEom @sovrasov Your comments were reflected
@negvet I'm currently facing issues with running unit-tests, I'll make a separate PR with unit tests for this

[negvet]

LGTM, please check pre-commits

[codecov-commenter]

Codecov Report

Base: 77.80% // Head: 77.76% // Decreases project coverage by -0.05% ⚠️

Coverage data is based on head (ef21b8c) compared to base (e5dd967).
Patch coverage: 29.50% of modified lines in pull request are covered.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1708      +/-   ##
===========================================
- Coverage    77.80%   77.76%   -0.05%     
===========================================
  Files          469      469              
  Lines        32800    32829      +29     
===========================================
+ Hits         25521    25528       +7     
- Misses        7279     7301      +22     

Impacted Files	Coverage Δ
otx/algorithms/common/tasks/training_base.py	50.72% <0.00%> (-0.75%)	⬇️
...x/mpa/modules/models/classifiers/sam_classifier.py	79.71% <0.00%> (-0.78%)	⬇️
...a/modules/models/detectors/custom_atss_detector.py	61.11% <0.00%> (-1.75%)	⬇️
...dules/models/detectors/custom_maskrcnn_detector.py	36.00% <0.00%> (-0.99%)	⬇️
...s/models/detectors/custom_single_stage_detector.py	27.65% <0.00%> (-0.61%)	⬇️
.../modules/models/detectors/custom_yolox_detector.py	32.92% <0.00%> (-0.83%)	⬇️
otx/algorithms/classification/tasks/openvino.py	86.66% <16.66%> (-1.23%)	⬇️
otx/algorithms/action/tasks/inference.py	88.38% <66.66%> (-0.29%)	⬇️
otx/algorithms/anomaly/tasks/inference.py	82.77% <66.66%> (-0.37%)	⬇️
otx/algorithms/segmentation/tasks/inference.py	88.96% <66.66%> (-0.98%)	⬇️
... and 8 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[GalyaZalesskaya]

@JihwanEom @sovrasov @negvet Tests were fixed, can you review again please?

[goodsong81]

Thank you!

[harimkang]

@GalyaZalesskaya I have a suspicion that instance-segmentation export is not working since this PR was merged.
Can you check it out?
i follow below steps:

otx train Custom_Counting_Instance_Segmentation_MaskRCNN_ResNet50 --train-data-roots tests/assets/car_tree_bug --val-data-roots tests/assets/car_tree_bug
cd otx-workspace-INSTANCE_SEGMENTATION
otx export

i got runtime error
RuntimeError: number of output names provided (5) exceeded number of outputs (3)