Update nonlocal readme in TSM and I3D (#131)

configs/recognition/i3d/README.md

 # I3D
 
+## Introduction
+```
+@inproceedings{inproceedings,
+  author = {Carreira, J. and Zisserman, Andrew},
+  year = {2017},
+  month = {07},
+  pages = {4724-4733},
+  title = {Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset},
+  doi = {10.1109/CVPR.2017.502}
+}
+
+@article{NonLocal2018,
+  author =   {Xiaolong Wang and Ross Girshick and Abhinav Gupta and Kaiming He},
+  title =    {Non-local Neural Networks},
+  journal =  {CVPR},
+  year =     {2018}
+}
+```
+
 ## Model Zoo
 
 ### Kinetics-400
@@ -12,6 +31,9 @@
 |[i3d_r50_dense_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_dense_32x2x1_100e_kinetics400_rgb.py) |short-side 256|8| ResNet50| ImageNet|73.48|91.00|x| 5170| [ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_dense_256p_32x2x1_100e_kinetics400_rgb/i3d_r50_dense_256p_32x2x1_100e_kinetics400_rgb_20200725-24eb54cc.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_dense_256p_32x2x1_100e_kinetics400_rgb/20200725_031604.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_dense_256p_32x2x1_100e_kinetics400_rgb/20200725_031604.log.json)|
 |[i3d_r50_fast_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_fast_32x2x1_100e_kinetics400_rgb.py) |340x256|8| ResNet50 |ImageNet|72.32|90.72|1.8 (320x3 frames)| 5170| [ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_fast_32x2x1_100e_kinetics400_rgb/i3d_r50_fast_32x2x1_100e_kinetics400_rgb_20200612-000e4d2a.pth) | [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_fast_32x2x1_100e_kinetics400_rgb/20200612_233836.log)| [json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_fast_32x2x1_100e_kinetics400_rgb/20200612_233836.log.json)|
 |[i3d_r50_fast_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_fast_32x2x1_100e_kinetics400_rgb.py) |short-side 256|8| ResNet50| ImageNet|73.24|90.99|x| 5170| [ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_fast_256p_32x2x1_100e_kinetics400_rgb/i3d_r50_fast_256p_32x2x1_100e_kinetics400_rgb_20200817-4e90d1d5.pth)| [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_fast_256p_32x2x1_100e_kinetics400_rgb/20200725_031457.log) | [json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_r50_fast_256p_32x2x1_100e_kinetics400_rgb/20200725_031457.log.json) |
+|[i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb.py)|short-side 256p|8x4| ResNet50 |ImageNet|74.71|91.81|x|6438|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb_20200813-6e6aef1b.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb/20200813_034054.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb/20200813_034054.log.json)|
+|[i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb.py)|short-side 256p|8x4| ResNet50 |ImageNet|73.37|91.26|x|4944|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb_20200815-17f84aa2.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb/20200813_034909.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb/20200813_034909.log.json)|
+|[i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb.py)|short-side 256p|8x4| ResNet50 |ImageNet|73.92|91.59|x|4832|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb_20200814-7c30d5bb.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb/20200814_044208.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/i3d/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb/20200814_044208.log.json)|
 
 Notes:
 1. The **gpus** indicates the number of gpu we used to get the checkpoint. It is noteworthy that the configs we provide are used for 8 gpus as default.

configs/recognition/i3d/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb.py

+# model settings
+model = dict(
+    type='Recognizer3D',
+    backbone=dict(
+        type='ResNet3d',
+        pretrained2d=True,
+        pretrained='torchvision://resnet50',
+        depth=50,
+        conv_cfg=dict(type='Conv3d'),
+        norm_eval=False,
+        inflate=((1, 1, 1), (1, 0, 1, 0), (1, 0, 1, 0, 1, 0), (0, 1, 0)),
+        non_local=((0, 0, 0), (0, 1, 0, 1), (0, 1, 0, 1, 0, 1), (0, 0, 0)),
+        non_local_cfg=dict(
+            sub_sample=True,
+            use_scale=False,
+            norm_cfg=dict(type='BN3d', requires_grad=True),
+            mode='dot_product'),
+        zero_init_residual=False),
+    cls_head=dict(
+        type='I3DHead',
+        num_classes=400,
+        in_channels=2048,
+        spatial_type='avg',
+        dropout_ratio=0.5,
+        init_std=0.01))
+# model training and testing settings
+train_cfg = None
+test_cfg = dict(average_clips=None)
+# dataset settings
+dataset_type = 'RawframeDataset'
+data_root = 'data/kinetics400/rawframes_train'
+data_root_val = 'data/kinetics400/rawframes_val'
+ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
+ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
+train_pipeline = [
+    dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(
+        type='MultiScaleCrop',
+        input_size=224,
+        scales=(1, 0.8),
+        random_crop=False,
+        max_wh_scale_gap=0),
+    dict(type='Resize', scale=(224, 224), keep_ratio=False),
+    dict(type='Flip', flip_ratio=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCTHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs', 'label'])
+]
+val_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=32,
+        frame_interval=2,
+        num_clips=1,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCTHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+test_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=32,
+        frame_interval=2,
+        num_clips=10,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='ThreeCrop', crop_size=256),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCTHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+data = dict(
+    videos_per_gpu=8,
+    workers_per_gpu=4,
+    train=dict(
+        type=dataset_type,
+        ann_file=ann_file_train,
+        data_prefix=data_root,
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        ann_file=ann_file_val,
+        data_prefix=data_root_val,
+        pipeline=val_pipeline),
+    test=dict(
+        type=dataset_type,
+        ann_file=ann_file_val,
+        data_prefix=data_root_val,
+        pipeline=test_pipeline))
+# optimizer
+optimizer = dict(
+    type='SGD', lr=0.01, momentum=0.9,
+    weight_decay=0.0001)  # this lr is used for 8 gpus
+optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
+# learning policy
+lr_config = dict(policy='step', step=[40, 80])
+total_epochs = 100
+checkpoint_config = dict(interval=5)
+evaluation = dict(
+    interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(
+    interval=20,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook'),
+    ])
+# runtime settings
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/i3d_nl_dot_product_r50_32x2x1_100e_kinetics400_rgb/'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]

configs/recognition/i3d/i3d_nl_r50_32x2x1_100e_kinetics400_rgb.py → configs/recognition/i3d/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb.py

@@ -122,7 +122,7 @@ log_config = dict(
 # runtime settings
 dist_params = dict(backend='nccl')
 log_level = 'INFO'
-work_dir = './work_dirs/i3d_nl_r50_32x2x1_100e_kinetics400_rgb/'
+work_dir = './work_dirs/i3d_nl_embedded_gaussian_r50_32x2x1_100e_kinetics400_rgb/'  # noqa: E501
 load_from = None
 resume_from = None
 workflow = [('train', 1)]

configs/recognition/i3d/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb.py

+# model settings
+model = dict(
+    type='Recognizer3D',
+    backbone=dict(
+        type='ResNet3d',
+        pretrained2d=True,
+        pretrained='torchvision://resnet50',
+        depth=50,
+        conv_cfg=dict(type='Conv3d'),
+        norm_eval=False,
+        inflate=((1, 1, 1), (1, 0, 1, 0), (1, 0, 1, 0, 1, 0), (0, 1, 0)),
+        non_local=((0, 0, 0), (0, 1, 0, 1), (0, 1, 0, 1, 0, 1), (0, 0, 0)),
+        non_local_cfg=dict(
+            sub_sample=True,
+            use_scale=False,
+            norm_cfg=dict(type='BN3d', requires_grad=True),
+            mode='gaussian'),
+        zero_init_residual=False),
+    cls_head=dict(
+        type='I3DHead',
+        num_classes=400,
+        in_channels=2048,
+        spatial_type='avg',
+        dropout_ratio=0.5,
+        init_std=0.01))
+# model training and testing settings
+train_cfg = None
+test_cfg = dict(average_clips=None)
+# dataset settings
+dataset_type = 'RawframeDataset'
+data_root = 'data/kinetics400/rawframes_train'
+data_root_val = 'data/kinetics400/rawframes_val'
+ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
+ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
+train_pipeline = [
+    dict(type='SampleFrames', clip_len=32, frame_interval=2, num_clips=1),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(
+        type='MultiScaleCrop',
+        input_size=224,
+        scales=(1, 0.8),
+        random_crop=False,
+        max_wh_scale_gap=0),
+    dict(type='Resize', scale=(224, 224), keep_ratio=False),
+    dict(type='Flip', flip_ratio=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCTHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs', 'label'])
+]
+val_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=32,
+        frame_interval=2,
+        num_clips=1,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCTHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+test_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=32,
+        frame_interval=2,
+        num_clips=10,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='ThreeCrop', crop_size=256),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCTHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+data = dict(
+    videos_per_gpu=8,
+    workers_per_gpu=4,
+    train=dict(
+        type=dataset_type,
+        ann_file=ann_file_train,
+        data_prefix=data_root,
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        ann_file=ann_file_val,
+        data_prefix=data_root_val,
+        pipeline=val_pipeline),
+    test=dict(
+        type=dataset_type,
+        ann_file=ann_file_val,
+        data_prefix=data_root_val,
+        pipeline=test_pipeline))
+# optimizer
+optimizer = dict(
+    type='SGD', lr=0.01, momentum=0.9,
+    weight_decay=0.0001)  # this lr is used for 8 gpus
+optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
+# learning policy
+lr_config = dict(policy='step', step=[40, 80])
+total_epochs = 100
+checkpoint_config = dict(interval=5)
+evaluation = dict(
+    interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(
+    interval=20,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook'),
+    ])
+# runtime settings
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/i3d_nl_gaussian_r50_32x2x1_100e_kinetics400_rgb/'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]

configs/recognition/tsm/README.md

 # TSM
 
+## Introduction
+```
+@inproceedings{lin2019tsm,
+  title={TSM: Temporal Shift Module for Efficient Video Understanding},
+  author={Lin, Ji and Gan, Chuang and Han, Song},
+  booktitle={Proceedings of the IEEE International Conference on Computer Vision},
+  year={2019}
+}
+
+@article{NonLocal2018,
+  author =   {Xiaolong Wang and Ross Girshick and Abhinav Gupta and Kaiming He},
+  title =    {Non-local Neural Networks},
+  journal =  {CVPR},
+  year =     {2018}
+}
+```
+
 ## Model Zoo
 
 ### Kinetics-400
@@ -13,6 +30,9 @@
 |[tsm_r50_dense_1x1x8_100e_kinetics400_rgb](/configs/recognition/tsm/tsm_r50_dense_1x1x8_100e_kinetics400_rgb.py) |short-side 256|8| ResNet50 | ImageNet|73.38|91.02|x|x|x|7079|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_dense_256p_1x1x8_100e_kinetics400_rgb/tsm_r50_dense_256p_1x1x8_100e_kinetics400_rgb_20200727-e1e0c785.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_dense_256p_1x1x8_100e_kinetics400_rgb/20200725_032043.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_dense_256p_1x1x8_100e_kinetics400_rgb/20200725_032043.log.json)|
 |[tsm_r50_1x1x16_50e_kinetics400_rgb](/configs/recognition/tsm/tsm_r50_1x1x16_50e_kinetics400_rgb.py) |340x256|8| ResNet50| ImageNet |71.69|90.4|[70.67](https://github.com/mit-han-lab/temporal-shift-module/blob/8d53d6fda40bea2f1b37a6095279c4b454d672bd/scripts/train_tsm_kinetics_rgb_16f.sh)|[89.98](https://github.com/mit-han-lab/temporal-shift-module/blob/8d53d6fda40bea2f1b37a6095279c4b454d672bd/scripts/train_tsm_kinetics_rgb_16f.sh)|47.0 (16x1 frames)| 10404  | [ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_1x1x16_50e_kinetics400_rgb/tsm_r50_1x1x16_50e_kinetics400_rgb_20200607-f731bffc.pth) | [log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_1x1x16_50e_kinetics400_rgb/20200607_221310.log)| [json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_1x1x16_50e_kinetics400_rgb/20200607_221310.log.json)|
 |[tsm_r50_1x1x16_50e_kinetics400_rgb](/configs/recognition/tsm/tsm_r50_1x1x16_50e_kinetics400_rgb.py) |short-side 256|8| ResNet50| ImageNet |72.01|90.57|x|x|x|10398|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_256p_1x1x16_50e_kinetics400_rgb/tsm_r50_256p_1x1x16_50e_kinetics400_rgb_20200727-b414aa3c.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_256p_1x1x16_50e_kinetics400_rgb/20200725_031232.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_r50_256p_1x1x16_50e_kinetics400_rgb/20200725_031232.log.json)|
+|[tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb](/configs/recognition/tsm/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb.py)|short-side 320|8x4| ResNet50| ImageNet |72.03|90.25|71.81|90.36|x|8931|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb_20200724-f00f1336.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb/20200724_120023.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb/20200724_120023.log.json)|
+|[tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb](/configs/recognition/tsm/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb.py)|short-side 320|8x4| ResNet50| ImageNet |70.70|89.90|x|x|x|10125|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb_20200816-b93fd297.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb/20200815_210253.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb/20200815_210253.log.json)|
+|[tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb](/configs/recognition/tsm/tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb.py)|short-side 320|8x4|ResNet50| ImageNet |71.60|90.34|x|x|x|8358|[ckpt](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb/tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb_20200724-d8ad84d2.pth)|[log](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb/20200723_220442.log)|[json](https://openmmlab.oss-accelerate.aliyuncs.com/mmaction/recognition/tsm/tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb/20200723_220442.log.json)|
 
 ### Something-Something V1
 

configs/recognition/tsm/tsm_nl_dot_product_r50_1x1x8_50e_kinetics400_rgb.py

+# model settings
+model = dict(
+    type='Recognizer2D',
+    backbone=dict(
+        type='ResNetTSM',
+        pretrained='torchvision://resnet50',
+        depth=50,
+        norm_eval=False,
+        non_local=((0, 0, 0), (1, 0, 1, 0), (1, 0, 1, 0, 1, 0), (0, 0, 0)),
+        non_local_cfg=dict(
+            sub_sample=True,
+            use_scale=False,
+            norm_cfg=dict(type='BN3d', requires_grad=True),
+            mode='dot_product'),
+        shift_div=8),
+    cls_head=dict(
+        type='TSMHead',
+        num_classes=400,
+        in_channels=2048,
+        spatial_type='avg',
+        consensus=dict(type='AvgConsensus', dim=1),
+        dropout_ratio=0.5,
+        init_std=0.001,
+        is_shift=True))
+# model training and testing settings
+train_cfg = None
+test_cfg = dict(average_clips=None)
+# dataset settings
+dataset_type = 'RawframeDataset'
+data_root = 'data/kinetics400/rawframes_train'
+data_root_val = 'data/kinetics400/rawframes_val'
+ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
+ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
+train_pipeline = [
+    dict(type='SampleFrames', clip_len=1, frame_interval=1, num_clips=8),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(
+        type='MultiScaleCrop',
+        input_size=224,
+        scales=(1, 0.875, 0.75, 0.66),
+        random_crop=False,
+        max_wh_scale_gap=1,
+        num_fixed_crops=13),
+    dict(type='Resize', scale=(224, 224), keep_ratio=False),
+    dict(type='Flip', flip_ratio=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs', 'label'])
+]
+val_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=1,
+        frame_interval=1,
+        num_clips=8,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+test_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=1,
+        frame_interval=1,
+        num_clips=8,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+data = dict(
+    videos_per_gpu=8,
+    workers_per_gpu=4,
+    train=dict(
+        type=dataset_type,
+        ann_file=ann_file_train,
+        data_prefix=data_root,
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        ann_file=ann_file_val,
+        data_prefix=data_root_val,
+        pipeline=val_pipeline),
+    test=dict(
+        type=dataset_type,
+        ann_file=ann_file_test,
+        data_prefix=data_root_val,
+        pipeline=test_pipeline))
+# optimizer
+optimizer = dict(
+    type='SGD',
+    constructor='TSMOptimizerConstructor',
+    paramwise_cfg=dict(fc_lr5=True),
+    lr=0.01,  # this lr is used for 8 gpus
+    momentum=0.9,
+    weight_decay=0.0001)
+optimizer_config = dict(grad_clip=dict(max_norm=20, norm_type=2))
+# learning policy
+lr_config = dict(policy='step', step=[20, 40])
+total_epochs = 50
+checkpoint_config = dict(interval=1)
+evaluation = dict(
+    interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(
+    interval=20,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook'),
+    ])
+# runtime settings
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb/'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]

configs/recognition/tsm/tsm_nl_r50_1x1x8_50e_kinetics400_rgb.py → configs/recognition/tsm/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb.py

@@ -126,7 +126,7 @@ log_config = dict(
 # runtime settings
 dist_params = dict(backend='nccl')
 log_level = 'INFO'
-work_dir = './work_dirs/tsm_nl_r50_1x1x8_50e_kinetics400_rgb/'
+work_dir = './work_dirs/tsm_nl_embedded_gaussian_r50_1x1x8_50e_kinetics400_rgb/'  # noqa: E501
 load_from = None
 resume_from = None
 workflow = [('train', 1)]

configs/recognition/tsm/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb.py

+# model settings
+model = dict(
+    type='Recognizer2D',
+    backbone=dict(
+        type='ResNetTSM',
+        pretrained='torchvision://resnet50',
+        depth=50,
+        norm_eval=False,
+        non_local=((0, 0, 0), (1, 0, 1, 0), (1, 0, 1, 0, 1, 0), (0, 0, 0)),
+        non_local_cfg=dict(
+            sub_sample=True,
+            use_scale=False,
+            norm_cfg=dict(type='BN3d', requires_grad=True),
+            mode='gaussian'),
+        shift_div=8),
+    cls_head=dict(
+        type='TSMHead',
+        num_classes=400,
+        in_channels=2048,
+        spatial_type='avg',
+        consensus=dict(type='AvgConsensus', dim=1),
+        dropout_ratio=0.5,
+        init_std=0.001,
+        is_shift=True))
+# model training and testing settings
+train_cfg = None
+test_cfg = dict(average_clips=None)
+# dataset settings
+dataset_type = 'RawframeDataset'
+data_root = 'data/kinetics400/rawframes_train'
+data_root_val = 'data/kinetics400/rawframes_val'
+ann_file_train = 'data/kinetics400/kinetics400_train_list_rawframes.txt'
+ann_file_val = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+ann_file_test = 'data/kinetics400/kinetics400_val_list_rawframes.txt'
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)
+train_pipeline = [
+    dict(type='SampleFrames', clip_len=1, frame_interval=1, num_clips=8),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(
+        type='MultiScaleCrop',
+        input_size=224,
+        scales=(1, 0.875, 0.75, 0.66),
+        random_crop=False,
+        max_wh_scale_gap=1,
+        num_fixed_crops=13),
+    dict(type='Resize', scale=(224, 224), keep_ratio=False),
+    dict(type='Flip', flip_ratio=0.5),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs', 'label'])
+]
+val_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=1,
+        frame_interval=1,
+        num_clips=8,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+test_pipeline = [
+    dict(
+        type='SampleFrames',
+        clip_len=1,
+        frame_interval=1,
+        num_clips=8,
+        test_mode=True),
+    dict(type='RawFrameDecode'),
+    dict(type='Resize', scale=(-1, 256)),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='Flip', flip_ratio=0),
+    dict(type='Normalize', **img_norm_cfg),
+    dict(type='FormatShape', input_format='NCHW'),
+    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['imgs'])
+]
+data = dict(
+    videos_per_gpu=8,
+    workers_per_gpu=4,
+    train=dict(
+        type=dataset_type,
+        ann_file=ann_file_train,
+        data_prefix=data_root,
+        pipeline=train_pipeline),
+    val=dict(
+        type=dataset_type,
+        ann_file=ann_file_val,
+        data_prefix=data_root_val,
+        pipeline=val_pipeline),
+    test=dict(
+        type=dataset_type,
+        ann_file=ann_file_test,
+        data_prefix=data_root_val,
+        pipeline=test_pipeline))
+# optimizer
+optimizer = dict(
+    type='SGD',
+    constructor='TSMOptimizerConstructor',
+    paramwise_cfg=dict(fc_lr5=True),
+    lr=0.01,  # this lr is used for 8 gpus
+    momentum=0.9,
+    weight_decay=0.0001)
+optimizer_config = dict(grad_clip=dict(max_norm=20, norm_type=2))
+# learning policy
+lr_config = dict(policy='step', step=[20, 40])
+total_epochs = 50
+checkpoint_config = dict(interval=1)
+evaluation = dict(
+    interval=5, metrics=['top_k_accuracy', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(
+    interval=20,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook'),
+    ])
+# runtime settings
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/tsm_nl_gaussian_r50_1x1x8_50e_kinetics400_rgb/'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]