[Fix] Fix hvu object numclasses (open-mmlab#307)

kennymckormick · web-flow · commit 4dfae3a5eddd · 2020-11-11T15:56:19.000+08:00
* resolve comments

* update changelog

* fix config

* fix typo (action model is trained with the right numclasses)

* update influenced model links

* fix bug

* remove file mis-added

* update changelog
diff --git a/configs/recognition/tsn/README.md b/configs/recognition/tsn/README.md
@@ -129,7 +129,7 @@ In data benchmark, we compare:
 | :----------------------------------------------------------: | :----------: | :------------: | :--: | :------: | :------: | :--: | :-------: | :-------------: | :----------------------------------------------------------: | :----------------------------------------------------------: | :----------------------------------------------------------: |
 | [tsn_r18_1x1x8_100e_hvu_action_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_action_rgb.py) |    action    | short-side 256 | 8x2  | ResNet18 | ImageNet | 57.5 |   51.8    |      53.5       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/action/tsn_r18_1x1x8_100e_hvu_action_rgb_20201027-011b282b.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/action/tsn_r18_1x1x8_100e_hvu_action_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/action/tsn_r18_1x1x8_100e_hvu_action_rgb_20201027.json) |
 | [tsn_r18_1x1x8_100e_hvu_scene_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_scene_rgb.py) |    scene     | short-side 256 |  8   | ResNet18 | ImageNet | 55.2 |   55.8    |      57.2       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/scene/tsn_r18_1x1x8_100e_hvu_scene_rgb_20201027-00e5748d.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/scene/tsn_r18_1x1x8_100e_hvu_scene_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/scene/tsn_r18_1x1x8_100e_hvu_scene_rgb_20201027.json) |
-| [tsn_r18_1x1x8_100e_hvu_object_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_object_rgb.py) |    object    | short-side 256 |  8   | ResNet18 | ImageNet | 45.7 |   34.2    |      35.1       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/object/tsn_r18_1x1x8_100e_hvu_object_rgb_20201027-ec8e1083.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/object/tsn_r18_1x1x8_100e_hvu_object_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/object/tsn_r18_1x1x8_100e_hvu_object_rgb_20201027.json) |
+| [tsn_r18_1x1x8_100e_hvu_object_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_object_rgb.py) |    object    | short-side 256 |  8   | ResNet18 | ImageNet | 45.7 |   34.2    |      35.1       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/object/tsn_r18_1x1x8_100e_hvu_object_rgb_20201102-24a22f30.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/object/tsn_r18_1x1x8_100e_hvu_object_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/object/tsn_r18_1x1x8_100e_hvu_object_rgb_20201027.json) |
 | [tsn_r18_1x1x8_100e_hvu_event_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_event_rgb.py) |    event     | short-side 256 |  8   | ResNet18 | ImageNet | 63.7 |   38.5    |      39.8       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/event/tsn_r18_1x1x8_100e_hvu_event_rgb_20201027-dea8cd71.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/event/tsn_r18_1x1x8_100e_hvu_event_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/event/tsn_r18_1x1x8_100e_hvu_event_rgb_20201027.json) |
 | [tsn_r18_1x1x8_100e_hvu_concept_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_concept_rgb.py) |   concept    | short-side 256 |  8   | ResNet18 | ImageNet | 47.5 |   26.1    |      27.3       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/concept/tsn_r18_1x1x8_100e_hvu_concept_rgb_20201027-fc1dd8e3.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/concept/tsn_r18_1x1x8_100e_hvu_concept_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/concept/tsn_r18_1x1x8_100e_hvu_concept_rgb_20201027.json) |
 | [tsn_r18_1x1x8_100e_hvu_attribute_rgb](/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_attribute_rgb.py) |  attribute   | short-side 256 |  8   | ResNet18 | ImageNet | 46.1 |   33.6    |      34.9       | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/attribute/tsn_r18_1x1x8_100e_hvu_attribute_rgb_20201027-0b3b49d2.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/attribute/tsn_r18_1x1x8_100e_hvu_attribute_rgb_20201027.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/hvu/attribute/tsn_r18_1x1x8_100e_hvu_attribute_rgb_20201027.json) |
diff --git a/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_action_rgb.py b/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_action_rgb.py
@@ -1,6 +1,6 @@
 # model settings
 category_nums = dict(
-    action=729, attribute=117, concept=291, event=69, object=1679, scene=248)
+    action=739, attribute=117, concept=291, event=69, object=1678, scene=248)
 target_cate = 'action'
 
 model = dict(
diff --git a/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_attribute_rgb.py b/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_attribute_rgb.py
@@ -1,6 +1,6 @@
 # model settings
 category_nums = dict(
-    action=729, attribute=117, concept=291, event=69, object=1679, scene=248)
+    action=739, attribute=117, concept=291, event=69, object=1678, scene=248)
 target_cate = 'attribute'
 
 model = dict(
diff --git a/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_concept_rgb.py b/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_concept_rgb.py
@@ -1,6 +1,6 @@
 # model settings
 category_nums = dict(
-    action=729, attribute=117, concept=291, event=69, object=1679, scene=248)
+    action=739, attribute=117, concept=291, event=69, object=1678, scene=248)
 target_cate = 'object'
 
 model = dict(
diff --git a/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_event_rgb.py b/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_event_rgb.py
@@ -1,6 +1,6 @@
 # model settings
 category_nums = dict(
-    action=729, attribute=117, concept=291, event=69, object=1679, scene=248)
+    action=739, attribute=117, concept=291, event=69, object=1678, scene=248)
 target_cate = 'event'
 
 model = dict(
diff --git a/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_object_rgb.py b/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_object_rgb.py
@@ -1,6 +1,6 @@
 # model settings
 category_nums = dict(
-    action=729, attribute=117, concept=291, event=69, object=1679, scene=248)
+    action=739, attribute=117, concept=291, event=69, object=1678, scene=248)
 target_cate = 'object'
 
 model = dict(
diff --git a/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_scene_rgb.py b/configs/recognition/tsn/hvu/tsn_r18_1x1x8_100e_hvu_scene_rgb.py
@@ -1,6 +1,6 @@
 # model settings
 category_nums = dict(
-    action=729, attribute=117, concept=291, event=69, object=1679, scene=248)
+    action=739, attribute=117, concept=291, event=69, object=1678, scene=248)
 target_cate = 'scene'
 
 model = dict(
diff --git a/docs/changelog.md b/docs/changelog.md
@@ -1,6 +1,6 @@
 ## Changelog
 
-###v 0.9.0 (master)
+### 0.9.0 (master)
 
 **Highlights**
 
@@ -14,6 +14,7 @@
 
 **Bug Fixes**
 - Fix a bug in BaseDataset when `data_prefix` is None ([#314](https://github.com/open-mmlab/mmaction2/pull/314))
+- Fix the bug of HVU object `num_classes` from 1679 to 1678 ([#307](https://github.com/open-mmlab/mmaction2/pull/307))
 
 **ModelZoo**
 
diff --git a/mmaction/models/losses/hvu_loss.py b/mmaction/models/losses/hvu_loss.py
@@ -14,7 +14,7 @@ class HVULoss(BaseWeightedLoss):
             this order. Default: ['action', 'attribute', 'concept', 'event',
             'object', 'scene'].
         category_nums (tuple[int]): Number of tags for each category. Default:
-            (739, 117, 291, 69, 1679, 248).
+            (739, 117, 291, 69, 1678, 248).
         category_loss_weights (tuple[float]): Loss weights of categories, it
             applies only if `loss_type == 'individual'`. The loss weights will
             be normalized so that the sum equals to 1, so that you can give any
@@ -34,7 +34,7 @@ class HVULoss(BaseWeightedLoss):
     def __init__(self,
                  categories=('action', 'attribute', 'concept', 'event',
                              'object', 'scene'),
-                 category_nums=(739, 117, 291, 69, 1679, 248),
+                 category_nums=(739, 117, 291, 69, 1678, 248),
                  category_loss_weights=(1, 1, 1, 1, 1, 1),
                  loss_type='all',
                  with_mask=False,
diff --git a/tests/test_data/test_dataset.py b/tests/test_data/test_dataset.py
@@ -141,7 +141,7 @@ def setup_class(cls):
             'action', 'attribute', 'concept', 'event', 'object', 'scene'
         ]
 
-        cls.hvu_category_nums = [739, 117, 291, 69, 1679, 248]
+        cls.hvu_category_nums = [739, 117, 291, 69, 1678, 248]
         cls.filename_tmpl = 'img_{:05d}.jpg'
 
     def test_hvu_dataset(self):

Original file line number	Diff line number	Diff line change
`@@ -141,7 +141,7 @@ def setup_class(cls):`
`141`	`141`	`'action', 'attribute', 'concept', 'event', 'object', 'scene'`
`142`	`142`	`]`
`143`	`143`
`144`		`- cls.hvu_category_nums = [739, 117, 291, 69, 1679, 248]`
	`144`	`+ cls.hvu_category_nums = [739, 117, 291, 69, 1678, 248]`
`145`	`145`	`cls.filename_tmpl = 'img_{:05d}.jpg'`
`146`	`146`
`147`	`147`	`def test_hvu_dataset(self):`