Upload 41 files

ntu120_xsub/b_2/b_2.py

@@ -0,0 +1,98 @@
+modality = 'b'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/b_2'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_1',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_3', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+    dict(type='GenSkeFeat', feats=['b']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['b']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['b']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+            dict(type='GenSkeFeat', feats=['b']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['b']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['b']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/b_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c91f52155366957683157822c456c44b4a3deb733dc795876b6839deca21d2e1
+size 37301266

ntu120_xsub/b_2/best_top1_acc_epoch_137.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88113dd9b6ac68935b01253b63c550a2b378690e523cec8b653641922366ebf5
+size 33657446

ntu120_xsub/bm/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:808bc13fd5cc33065f7f3dd71a41a516fe491dbf8efa237b28ccf1a52a8ede70
+size 37300681

ntu120_xsub/bm/best_top1_acc_epoch_148.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:234939d055d1be92185ce98bdbe070ffe0edc32e30a5d4f815d6f2f1f816a3e5
+size 33657318

ntu120_xsub/bm/bm.py

@@ -0,0 +1,95 @@
+modality = 'bm'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/bm'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_1',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_3', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+    dict(type='GenSkeFeat', feats=['bm']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['bm']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['bm']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+            dict(type='GenSkeFeat', feats=['bm']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['bm']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['bm']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/j_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a34a9013bd061a2951ab99ef97203f4bc55417d001e44f8eb633510da27a5355
+size 37307285

ntu120_xsub/j_1/best_top1_acc_epoch_141.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42bb9554a189ab3856da8a9ab92ac4cbf0dc307c868d242c1f77ca1853f1944a
+size 35500454

ntu120_xsub/j_1/j_1.py

@@ -0,0 +1,95 @@
+modality = 'j'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/j_1'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_2',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_2', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+    dict(type='GenSkeFeat', feats=['j']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['j']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['j']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+            dict(type='GenSkeFeat', feats=['j']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['j']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['j']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/j_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9b2d4e87efbcaac83761c906a3d859ff6da73e3c0dd1b4208aaf4e612c410c1
+size 37295569

ntu120_xsub/j_2/best_top1_acc_epoch_144.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e6768bd626c55f5ec17b8112126781bff1f5867a10e784b486af19ca14a4397
+size 35500646

ntu120_xsub/j_2/j_2.py

@@ -0,0 +1,98 @@
+modality = 'j'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/j_2'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_2',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_2', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+    dict(type='GenSkeFeat', feats=['j']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['j']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['j']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+            dict(type='GenSkeFeat', feats=['j']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['j']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['j']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/jm/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:505f1b3dd85f06c4bdd1d0866c6af9d702d798f16dcaa83e75bec75ad86c6f38
+size 37303711

ntu120_xsub/jm/best_top1_acc_epoch_148.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac6a32805a8fe2c3836908bf1a688ff02d2496bbb5dfc1e9c4ace45a9545de05
+size 35500518

ntu120_xsub/jm/jm.py

@@ -0,0 +1,95 @@
+modality = 'jm'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/jm'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_2',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_2', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+    dict(type='GenSkeFeat', feats=['jm']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['jm']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['jm']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Spatial_Flip', dataset='nturgb+d', p=0.5),
+            dict(type='GenSkeFeat', feats=['jm']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['jm']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['jm']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/k_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:646dafb8f731662a9d4f1bba1c7053d368747e04f22d0a41e8ee981891660b55
+size 37302175

ntu120_xsub/k_1/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94b94e18bb25674831aada966c264e8aab30379a7a32013366b0d5a4a0e48a66
+size 33657382

ntu120_xsub/k_1/k_1.py

@@ -0,0 +1,98 @@
+modality = 'k'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/k_1'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_1',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_1', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Part_Drop'),
+    dict(type='GenSkeFeat', feats=['k']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['k']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['k']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Part_Drop'),
+            dict(type='GenSkeFeat', feats=['k']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['k']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/hao.wang/reproducation/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['k']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/k_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fed796450407ad22e238196b030139878ec6e859724cb10cbd7b157646b2a004
+size 37302443

ntu120_xsub/k_2/best_top1_acc_epoch_142.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2958ce16195b792e225ded5daca8015ca112cca1f3c714002c62325ebe855e98
+size 33657446

ntu120_xsub/k_2/k_2.py

@@ -0,0 +1,98 @@
+modality = 'k'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/k_2'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_1',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_1', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Part_Drop'),
+    dict(type='GenSkeFeat', feats=['k']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['k']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['k']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Part_Drop'),
+            dict(type='GenSkeFeat', feats=['k']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['k']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file=
+        '/data1/medical/hao.wang/reproduction/hongda.liu/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['k']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

ntu120_xsub/km/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f971592e23236123fa23675831724d42380fd450d1d649a8df83004fc9bfa82
+size 37302765

ntu120_xsub/km/best_top1_acc_epoch_146.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54ba5233029dffa7e9371ae9cea15438ecf4bcd273f00012041126f9aae81eb0
+size 33657318

ntu120_xsub/km/km.py

@@ -0,0 +1,95 @@
+modality = 'km'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu120_xsub/km'
+model = dict(
+    type='RecognizerGCN_7_1_1',
+    backbone=dict(
+        type='GCN_7_1_1',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='nturgb+d',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead_7_4_1', num_classes=120, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    dict(type='Part_Drop'),
+    dict(type='GenSkeFeat', feats=['km']),
+    dict(type='UniformSampleDecode', clip_len=100),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['km']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='GenSkeFeat', feats=['km']),
+    dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+    dict(type='FormatGCNInput'),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+data = dict(
+    videos_per_gpu=16,
+    workers_per_gpu=4,
+    test_dataloader=dict(videos_per_gpu=1),
+    train=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            dict(type='Part_Drop'),
+            dict(type='GenSkeFeat', feats=['km']),
+            dict(type='UniformSampleDecode', clip_len=100),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['km']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=1),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/nturgbd/ntu120_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='GenSkeFeat', feats=['km']),
+            dict(type='UniformSampleDecode', clip_len=100, num_clips=10),
+            dict(type='FormatGCNInput'),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='xsub_val'))
+optimizer = dict(
+    type='SGD', lr=0.05, momentum=0.9, weight_decay=0.0005, nesterov=True)
+optimizer_config = dict(grad_clip=None)
+lr_config = dict(policy='CosineAnnealing', min_lr=0, by_epoch=False)
+total_epochs = 150
+checkpoint_config = dict(interval=1)
+evaluation = dict(interval=1, metrics=['top_k_accuracy'])
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 2)

ntu120_xsub/ntu120_xsub_ensemble.py

@@ -0,0 +1,65 @@
+from mmcv import load
+import sys
+# Note: please adjust the relative path according to the actual situation.
+sys.path.append('../..')
+from protogcn.smp import *
+
+
+j_1 = load('j_1/best_pred.pkl')
+b_1 = load('b_1/best_pred.pkl')
+k_1 = load('k_1/best_pred.pkl')
+j_2 = load('j_2/best_pred.pkl')
+b_2 = load('b_2/best_pred.pkl')
+k_2 = load('k_2/best_pred.pkl')
+jm = load('jm/best_pred.pkl')
+bm = load('bm/best_pred.pkl')
+km = load('km/best_pred.pkl')
+label = load_label('/data/nturgbd/ntu120_3danno.pkl', 'xsub_val')
+
+
+"""
+***************
+InfoGCN  v0:
+j jm b bm k km
+2S:     89.75
+4S:     89.86
+6S:     90.44
+***************
+"""
+print('InfoGCN  v0:')
+print('j jm b bm k km')
+print('2S')
+fused = comb([j_1, b_1], [1, 1])
+print('Top-1', top1(fused, label))
+
+print('4S')
+fused = comb([j_1, b_1, jm, bm], [2, 2, 1, 1])
+print('Top-1', top1(fused, label))
+
+print('6S')
+fused = comb([j_1, b_1, k_1, jm, bm, km], [7, 7, 7, 3, 3, 3])
+print('Top-1', top1(fused, label))
+
+
+"""
+***************
+InfoGCN  v1:
+j j b b k k
+2S:     89.75
+4S:     90.43
+6S:     90.92
+***************
+"""
+print('InfoGCN  v1:')
+print('j j b b k k')
+print('2S')
+fused = comb([j_1, b_1], [1, 1])
+print('Top-1', top1(fused, label))
+
+print('4S')
+fused = comb([j_1, b_1, j_2, b_2], [4, 9, 4, 9])
+print('Top-1', top1(fused, label))
+
+print('6S')
+fused = comb([j_1, j_2, b_1, b_2, k_1, k_2], [4, 4, 9, 9, 5, 5])
+print('Top-1', top1(fused, label))