reduce num_iters, rename DetClassProbMapHook

otx/algorithms/detection/adapters/mmdet/hooks/__init__.py

@@ -3,6 +3,6 @@
 # SPDX-License-Identifier: Apache-2.0
 #
 
-from .det_class_probability_map_hook import DetClassProbabilityMap
+from .det_class_probability_map_hook import DetClassProbabilityMapHook
 
-__all__ = ["DetClassProbabilityMap"]
+__all__ = ["DetClassProbabilityMapHook"]

otx/algorithms/detection/adapters/mmdet/hooks/det_class_probability_map_hook.py

@@ -24,7 +24,8 @@
 # pylint: disable=too-many-locals
 
 
-class DetClassProbabilityMap(BaseRecordingForwardHook):
+
+class DetClassProbabilityMapHook(BaseRecordingForwardHook):
     """Saliency map hook for object detection models."""
 
     def __init__(self, module: torch.nn.Module) -> None:

otx/algorithms/detection/adapters/mmdet/models/detectors/custom_atss_detector.py

@@ -10,7 +10,7 @@
 from mmdet.models.detectors.atss import ATSS
 
 from otx.algorithms.detection.adapters.mmdet.hooks.det_class_probability_map_hook import (
-    DetClassProbabilityMap,
+    DetClassProbabilityMapHook,
 )
 from otx.mpa.deploy.utils import is_mmdeploy_enabled
 from otx.mpa.modules.hooks.recording_forward_hooks import FeatureVectorHook
@@ -97,7 +97,7 @@ def custom_atss__simple_test(ctx, self, img, img_metas, **kwargs):
         if ctx.cfg["dump_features"]:
             feature_vector = FeatureVectorHook.func(feat)
             cls_scores = outs[0]
-            saliency_map = DetClassProbabilityMap(self).func(feature_map=cls_scores, cls_scores_provided=True)
+            saliency_map = DetClassProbabilityMapHook(self).func(feature_map=cls_scores, cls_scores_provided=True)
             return (*bbox_results, feature_vector, saliency_map)
 
         return bbox_results

otx/algorithms/detection/adapters/mmdet/models/detectors/custom_single_stage_detector.py

@@ -10,7 +10,7 @@
 from mmdet.models.detectors.single_stage import SingleStageDetector
 
 from otx.algorithms.detection.adapters.mmdet.hooks.det_class_probability_map_hook import (
-    DetClassProbabilityMap,
+    DetClassProbabilityMapHook,
 )
 from otx.mpa.deploy.utils import is_mmdeploy_enabled
 from otx.mpa.modules.hooks.recording_forward_hooks import FeatureVectorHook
@@ -154,7 +154,7 @@ def custom_single_stage_detector__simple_test(ctx, self, img, img_metas, **kwarg
         if ctx.cfg["dump_features"]:
             feature_vector = FeatureVectorHook.func(feat)
             cls_scores = outs[0]
-            saliency_map = DetClassProbabilityMap(self).func(cls_scores, cls_scores_provided=True)
+            saliency_map = DetClassProbabilityMapHook(self).func(cls_scores, cls_scores_provided=True)
             return (*bbox_results, feature_vector, saliency_map)
 
         return bbox_results

otx/algorithms/detection/adapters/mmdet/models/detectors/custom_yolox_detector.py

@@ -10,7 +10,7 @@
 from mmdet.models.detectors.yolox import YOLOX
 
 from otx.algorithms.detection.adapters.mmdet.hooks.det_class_probability_map_hook import (
-    DetClassProbabilityMap,
+    DetClassProbabilityMapHook,
 )
 from otx.mpa.deploy.utils import is_mmdeploy_enabled
 from otx.mpa.modules.hooks.recording_forward_hooks import FeatureVectorHook
@@ -135,7 +135,7 @@ def custom_yolox__simple_test(ctx, self, img, img_metas, **kwargs):
         if ctx.cfg["dump_features"]:
             feature_vector = FeatureVectorHook.func(feat)
             cls_scores = outs[0]
-            saliency_map = DetClassProbabilityMap(self).func(cls_scores, cls_scores_provided=True)
+            saliency_map = DetClassProbabilityMapHook(self).func(cls_scores, cls_scores_provided=True)
             return (*bbox_results, feature_vector, saliency_map)
 
         return bbox_results

otx/mpa/det/explainer.py

@@ -14,7 +14,7 @@
 )
 from otx.algorithms.detection.adapters.mmdet.datasets import ImageTilingDataset
 from otx.algorithms.detection.adapters.mmdet.hooks.det_class_probability_map_hook import (
-    DetClassProbabilityMap,
+    DetClassProbabilityMapHook,
 )
 from otx.mpa.modules.hooks.recording_forward_hooks import (
     ActivationMapHook,
@@ -27,7 +27,7 @@
 
 logger = get_logger()
 EXPLAINER_HOOK_SELECTOR = {
-    "classwisesaliencymap": DetClassProbabilityMap,
+    "classwisesaliencymap": DetClassProbabilityMapHook,
     "eigencam": EigenCamHook,
     "activationmap": ActivationMapHook,
 }

otx/mpa/det/inferrer.py

@@ -19,7 +19,7 @@
 )
 from otx.algorithms.detection.adapters.mmdet.datasets import ImageTilingDataset
 from otx.algorithms.detection.adapters.mmdet.hooks.det_class_probability_map_hook import (
-    DetClassProbabilityMap,
+    DetClassProbabilityMapHook,
 )
 from otx.mpa.modules.hooks.recording_forward_hooks import (
     ActivationMapHook,
@@ -179,7 +179,7 @@ def infer(self, cfg, model_builder=None, eval=False, dump_features=False, dump_s
             if isinstance(raw_model, TwoStageDetector):
                 saliency_hook = ActivationMapHook(feature_model)
             else:
-                saliency_hook = DetClassProbabilityMap(feature_model)
+                saliency_hook = DetClassProbabilityMapHook(feature_model)
 
         eval_predictions = []
         with FeatureVectorHook(feature_model) if dump_features else nullcontext() as feature_vector_hook:

tests/e2e/test_api_xai_sanity.py

@@ -150,7 +150,7 @@ class TestOVDetXAIAPI(DetectionTaskAPIBase):
     def test_inference_xai(self):
         with tempfile.TemporaryDirectory() as temp_dir:
             hyper_parameters, model_template = self.setup_configurable_parameters(
-                DEFAULT_DET_TEMPLATE_DIR, num_iters=100
+                DEFAULT_DET_TEMPLATE_DIR, num_iters=15
             )
             detection_environment, dataset = self.init_environment(hyper_parameters, model_template, 10)
 

tests/test_suite/run_test_command.py

@@ -650,7 +650,7 @@ def xfail_templates(templates, xfail_template_ids_reasons):
     return xfailed_templates
 
 
-def otx_explain_testing(template, root, otx_dir, args):
+def otx_explain_testing(template, root, otx_dir, args, trained=False):
     template_work_dir = get_template_dir(template, root)
     if "RCNN" in template.model_template_id:
         test_algorithm = "ActivationMap"
@@ -683,8 +683,9 @@ def otx_explain_testing(template, root, otx_dir, args):
     ]
     check_run(command_line)
     assert os.path.exists(output_dir)
-    assert len(os.listdir(output_dir)) > 0
-    assert all([fname.split(".")[-1] == "tiff" for fname in os.listdir(output_dir)])
+    if trained:
+        assert len(os.listdir(output_dir)) > 0
+        assert all([fname.split(".")[-1] == "tiff" for fname in os.listdir(output_dir)])
 
 
 def otx_explain_testing_all_classes(template, root, otx_dir, args):
@@ -731,7 +732,7 @@ def otx_explain_testing_all_classes(template, root, otx_dir, args):
     assert all([fname.split(".")[-1] == "tiff" for fname in os.listdir(output_dir)])
 
 
-def otx_explain_testing_process_saliency_maps(template, root, otx_dir, args):
+def otx_explain_testing_process_saliency_maps(template, root, otx_dir, args, trained=False):
     template_work_dir = get_template_dir(template, root)
     if "RCNN" in template.model_template_id:
         test_algorithm = "ActivationMap"
@@ -765,11 +766,12 @@ def otx_explain_testing_process_saliency_maps(template, root, otx_dir, args):
     ]
     check_run(command_line)
     assert os.path.exists(output_dir)
-    assert len(os.listdir(output_dir)) > 0
-    assert all([fname.split(".")[-1] == "png" for fname in os.listdir(output_dir)])
+    if trained:
+        assert len(os.listdir(output_dir)) > 0
+        assert all([fname.split(".")[-1] == "png" for fname in os.listdir(output_dir)])
 
 
-def otx_explain_openvino_testing(template, root, otx_dir, args):
+def otx_explain_openvino_testing(template, root, otx_dir, args, trained=False):
     template_work_dir = get_template_dir(template, root)
     if "RCNN" in template.model_template_id:
         test_algorithm = "ActivationMap"
@@ -803,8 +805,9 @@ def otx_explain_openvino_testing(template, root, otx_dir, args):
     assert os.path.exists(f"{template_work_dir}/exported_{template.model_template_id}_w_features/openvino.xml")
     check_run(command_line)
     assert os.path.exists(output_dir)
-    assert len(os.listdir(output_dir)) > 0
-    assert all([fname.split(".")[-1] == "tiff" for fname in os.listdir(output_dir)])
+    if trained:
+        assert len(os.listdir(output_dir)) > 0
+        assert all([fname.split(".")[-1] == "tiff" for fname in os.listdir(output_dir)])
 
 
 def otx_explain_all_classes_openvino_testing(template, root, otx_dir, args):
@@ -852,7 +855,7 @@ def otx_explain_all_classes_openvino_testing(template, root, otx_dir, args):
     assert all([fname.split(".")[-1] == "tiff" for fname in os.listdir(output_dir)])
 
 
-def otx_explain_process_saliency_maps_openvino_testing(template, root, otx_dir, args):
+def otx_explain_process_saliency_maps_openvino_testing(template, root, otx_dir, args, trained=False):
     template_work_dir = get_template_dir(template, root)
     if "RCNN" in template.model_template_id:
         test_algorithm = "ActivationMap"
@@ -887,8 +890,9 @@ def otx_explain_process_saliency_maps_openvino_testing(template, root, otx_dir,
     assert os.path.exists(f"{template_work_dir}/exported_{template.model_template_id}_w_features/openvino.xml")
     check_run(command_line)
     assert os.path.exists(output_dir)
-    assert len(os.listdir(output_dir)) > 0
-    assert all([fname.split(".")[-1] == "png" for fname in os.listdir(output_dir)])
+    if trained:
+        assert len(os.listdir(output_dir)) > 0
+        assert all([fname.split(".")[-1] == "png" for fname in os.listdir(output_dir)])
 
 
 def otx_find_testing():

tests/unit/algorithms/detection/adapters/mmdet/hooks/test_det_saliency_map_hook.py → tests/unit/algorithms/detection/adapters/mmdet/hooks/test_class_probability_map_hook.py

@@ -1,8 +1,8 @@
 import pytest
 import torch
 
-from otx.algorithms.detection.adapters.mmdet.hooks.det_saliency_map_hook import (
-    DetSaliencyMapHook,
+from otx.algorithms.detection.adapters.mmdet.hooks.det_class_probability_map_hook import (
+    DetClassProbabilityMapHook,
 )
 from otx.algorithms.detection.adapters.mmdet.models.heads.custom_atss_head import (
     CustomATSSHead,
@@ -19,8 +19,8 @@
 from tests.test_suite.e2e_test_system import e2e_pytest_unit
 
 
-class TestDetSaliencyMapHook:
-    """Test class for DetSaliencyMapHook."""
+class TestDetClassProbabilityMapHook:
+    """Test class for DetClassProbabilityMapHook."""
 
     @pytest.fixture(autouse=True)
     def setup(self) -> None:
@@ -37,14 +37,14 @@ def forward(self, x):
                 return x
 
         self.module = _MockModule()
-        self.hook = DetSaliencyMapHook(self.module)
+        self.hook = DetClassProbabilityMapHook(self.module)
 
     @e2e_pytest_unit
     def test_func(self, mocker) -> None:
         """Test func function."""
 
         mocker.patch.object(
-            DetSaliencyMapHook, "_get_cls_scores_from_feature_map", return_value=[torch.randn(1, 3, 14, 14)]
+            DetClassProbabilityMapHook, "_get_cls_scores_from_feature_map", return_value=[torch.randn(1, 3, 14, 14)]
         )
         assert self.hook.func(torch.randn(1, 3, 14, 14)) is not None
 
@@ -53,14 +53,14 @@ def test_get_cls_scores_from_feature_map(self) -> None:
         """Test _get_cls_scores_from_feature_map function."""
 
         self.module.bbox_head = CustomATSSHead(num_classes=3, in_channels=64)
-        self.hook = DetSaliencyMapHook(self.module)
+        self.hook = DetClassProbabilityMapHook(self.module)
         assert self.hook._get_cls_scores_from_feature_map(torch.Tensor(1, 3, 64, 32, 32)) is not None
         self.module.bbox_head = CustomYOLOXHead(num_classes=3, in_channels=64)
-        self.hook = DetSaliencyMapHook(self.module)
+        self.hook = DetClassProbabilityMapHook(self.module)
         assert self.hook._get_cls_scores_from_feature_map(torch.Tensor(1, 3, 64, 32, 32)) is not None
         self.module.bbox_head = CustomVFNetHead(num_classes=3, in_channels=64)
         self.module.bbox_head.anchor_generator.num_base_anchors = 1
-        self.hook = DetSaliencyMapHook(self.module)
+        self.hook = DetClassProbabilityMapHook(self.module)
         assert self.hook._get_cls_scores_from_feature_map(torch.Tensor(1, 3, 64, 32, 32)) is not None
         self.module.bbox_head = CustomSSDHead(
             anchor_generator=dict(
@@ -71,10 +71,10 @@ def test_get_cls_scores_from_feature_map(self) -> None:
             ),
             act_cfg={},
         )
-        self.hook = DetSaliencyMapHook(self.module)
+        self.hook = DetClassProbabilityMapHook(self.module)
         assert self.hook._get_cls_scores_from_feature_map(torch.Tensor(1, 3, 512, 32, 32)) is not None
         self.module.bbox_head = torch.nn.Module()
         self.module.bbox_head.cls_out_channels = 3
-        self.hook = DetSaliencyMapHook(self.module)
+        self.hook = DetClassProbabilityMapHook(self.module)
         with pytest.raises(NotImplementedError):
             self.hook._get_cls_scores_from_feature_map(torch.Tensor(1, 3, 512, 32, 32))

tests/unit/algorithms/detection/test_xai_detection_validity.py

@@ -10,7 +10,7 @@
 from mmdet.models import build_detector
 
 from otx.algorithms.classification.tasks import ClassificationInferenceTask  # noqa
-from otx.algorithms.detection.adapters.mmdet.hooks import DetClassProbabilityMap
+from otx.algorithms.detection.adapters.mmdet.hooks import DetClassProbabilityMapHook
 from otx.cli.registry import Registry
 from otx.mpa.det.stage import DetectionStage  # noqa
 from otx.mpa.utils.config_utils import MPAConfig
@@ -61,7 +61,7 @@ def test_saliency_map_det(self, template):
         ]
         data = {"img_metas": [img_metas], "img": [img]}
 
-        with DetClassProbabilityMap(model) as det_hook:
+        with DetClassProbabilityMapHook(model) as det_hook:
             with torch.no_grad():
                 _ = model(return_loss=False, rescale=True, **data)
         saliency_maps = det_hook.records