Skip to content
This repository has been archived by the owner on Apr 17, 2023. It is now read-only.

Don't apply activations on export in classification #83

Merged
merged 2 commits into from
Nov 14, 2022

Conversation

sovrasov
Copy link
Contributor

The motivation is that d-o-r, POT, OpenVINO task and Model API classification components don't expect activated output from IR, so we had wrong confidence results of classification and simply wrong results of multilabel classification in OV tasks / POT. As soon as we remove d-o-r and related code, we can use activated output on the export sage again.

Copy link
Collaborator

@goodsong81 goodsong81 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Let's align outputs after DOR/MPA/OTX consolidation.

Copy link
Contributor

@JihwanEom JihwanEom left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@sovrasov sovrasov merged commit 3e2551b into develop Nov 14, 2022
@sovrasov sovrasov deleted the vs/fix_multilabel_export branch November 14, 2022 12:47
goodsong81 added a commit that referenced this pull request Dec 28, 2022
* fixed multilabel configs (#67)

* Tiling Module (#40)

* Feature/val batch and seed (#69)

* workaround bug (#70)

* Kp/devide runners (#71)

* move ema model to hook (#73)

* Appley release/OTE_0.3.0 changes (#77)

* Per-group softmax output (#74)

* Move drop_last in cls trainer.py (#79)

* Removed unnecessary mim version constraint for networkx package (#80)

* Revert "Removed unnecessary mim version constraint for networkx package (#80)" (#82)

* Don't apply activations on export in classification (#83)

* Delete unused code (#84)

* Replace current saliency map generation with Recipro-CAM for cls (#81)

* Class-wise saliency map generation for the detection task (#97)

* [XAI] hot-fix of error in Detection XAI support (#98)

* [XAI] hot-fix of error in Detection XAI support (#99)

* Replace only the first occurrence in the state_dict keys (#91)

* [OTE / XAI] Handle two stage detector in the inferrer.py (#104)

* [XAI] hot-fix of error in Detection XAI support (#98)

* [OTE / XAI][Develop] Handle two stage detector in the inferrer.py (#107)

* [XAI] hot-fix of error in Detection XAI support (#99)

* Merge OTE side XAI update to OTX (#109)

* Merge back releases/OTE_0.4.0 to develop (#116)

* [XAI] hot-fix of error in Detection XAI support (#99)

* [OTE / XAI] Handle two stage detector in the inferrer.py (#104)

* [XAI] hot-fix of error in Detection XAI support (#98)

* [Hot-fix] Fix zero-division error in one cycle lr scheduler in multilabel classification

* Fix extra activations when exporting nonlinear hierarchical head (#118)

* Fix get_train_data_cfg -> get_data_cfg

Signed-off-by: Songki Choi <songki.choi@intel.com>
Co-authored-by: Prokofiev Kirill <kirill.prokofiev@intel.com>
Co-authored-by: Eugene Liu <eugene.liu@intel.com>
Co-authored-by: Sungman Cho <sungman.cho@intel.com>
Co-authored-by: Jihwan Eom <jihwan.eom@intel.com>
Co-authored-by: Yunchu Lee <yunchu.lee@intel.com>
Co-authored-by: Vladislav Sovrasov <vladislav.sovrasov@intel.com>
Co-authored-by: Evgeny Tsykunov <e.tsykunov@gmail.com>
Co-authored-by: dongkwan-kim <dongkwan.kim@intel.com>
Co-authored-by: Adria Arrufat <adria.arrufat@intel.com>
Co-authored-by: Soobee Lee <soobee.lee@intel.com>
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants