Upload 41 files

ntu60_xsub/b_2/b_2.py

@@ -0,0 +1,95 @@
+modality = 'b'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_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='/data/lhd/pyskl_data/nturgbd/ntu60_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='/data/lhd/pyskl_data/nturgbd/ntu60_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='/data/lhd/pyskl_data/nturgbd/ntu60_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)

ntu60_xsub/b_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e189bfbc993f4600347b97a7d4f7f14ebc9e3048bd0035b585ddb728d923e872
+size 6636301

ntu60_xsub/b_2/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:057c434c0ae26e9045af62905e81531be88b16d3d1758bd629e1acab06137cba
+size 33472486

ntu60_xsub/bm/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95e56d00a9b50bd241bb59fed4cd1138b5dde74361abb93d631c2d8f8eec554a
+size 6638098

ntu60_xsub/bm/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2db5a0292c4aca801c35698f8451184939e22d95ab7529ee99d43c119db125c5
+size 33472486

ntu60_xsub/bm/bm.py

@@ -0,0 +1,95 @@
+modality = 'bm'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_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/ntu60_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/ntu60_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/ntu60_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)

ntu60_xsub/j_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03499b0dc1823e5c3612736cece778cb1e0d0ee4e4f9b07ce33295ad4f7db597
+size 6639657

ntu60_xsub/j_1/best_top1_acc_epoch_149.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdfb3cebf628c997921518f6ff897d8195a1e1310152c7ebd26613dff623dc5e
+size 35315686

ntu60_xsub/j_1/j_1.py

@@ -0,0 +1,93 @@
+modality = 'j'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    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/ntu60_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            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/ntu60_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/ntu60_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)

ntu60_xsub/j_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31c61d3aa107d57ebbf2f47a01e4e6a3ea64d5f57da42194d502f399e5d5c443
+size 6634538

ntu60_xsub/j_2/best_top1_acc_epoch_142.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00cb9b188ab8d6a0e66021762e1c3c400f04029b146ce55356bd9cdc9c5b6657
+size 35315686

ntu60_xsub/j_2/j_2.py

@@ -0,0 +1,93 @@
+modality = 'j'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    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/ntu60_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            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/ntu60_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/ntu60_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)

ntu60_xsub/jm/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:126dcb53123cdc8dd74798f1266a68f9d827e4793285241917935c16c6e07cca
+size 6639033

ntu60_xsub/jm/best_top1_acc_epoch_147.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01a10133fc05e8f77519bb20279b4652a3a1c5643a59e3c06b2a95a5da326ff3
+size 35315686

ntu60_xsub/jm/jm.py

@@ -0,0 +1,93 @@
+modality = 'jm'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_3danno.pkl'
+train_pipeline = [
+    dict(type='PreNormalize3D', align_spine=False),
+    dict(type='RandomRot', theta=0.2),
+    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/ntu60_3danno.pkl',
+        pipeline=[
+            dict(type='PreNormalize3D', align_spine=False),
+            dict(type='RandomRot', theta=0.2),
+            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/ntu60_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/ntu60_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)

ntu60_xsub/k_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d8866bcbda5640066221bb17ae67dfd76344717952c5e72b03e026671754491
+size 6638194

ntu60_xsub/k_1/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acff47d51c01a74d1c612d9881c1c950f676eb15999a4548e1f4a7fc2684aaf4
+size 33472486

ntu60_xsub/k_1/k_1.py

@@ -0,0 +1,95 @@
+modality = 'k'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_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=['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='/data/lhd/pyskl_data/nturgbd/ntu60_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=['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='/data/lhd/pyskl_data/nturgbd/ntu60_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='/data/lhd/pyskl_data/nturgbd/ntu60_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)

ntu60_xsub/k_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:357ba533d32664dbff563a10f48fac584b674bad805acfffbb1befc0007aa8c3
+size 6639038

ntu60_xsub/k_2/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba8b38496a25924d347aec21e4e7006f6ed19c0e4abc06443ee791da8aebf5fb
+size 33472486

ntu60_xsub/k_2/k_2.py

@@ -0,0 +1,95 @@
+modality = 'k'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_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=['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='/data/lhd/pyskl_data/nturgbd/ntu60_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=['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='/data/lhd/pyskl_data/nturgbd/ntu60_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='/data/lhd/pyskl_data/nturgbd/ntu60_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)

ntu60_xsub/km/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c69dfb5cee3f4e88b2fba2d54b53a5036d08c730d85e7ece32e380de53fad80
+size 6635960

ntu60_xsub/km/best_top1_acc_epoch_147.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e140e9b4e56001408804f205b9e65405d0deb738936f6d56ed8a58ad75971847
+size 33472486

ntu60_xsub/km/km.py

@@ -0,0 +1,95 @@
+modality = 'km'
+graph = 'nturgb+d'
+work_dir = './work_dirs/test_prototype/ntu60_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=60, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/nturgbd/ntu60_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=['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/ntu60_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=['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/ntu60_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/ntu60_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=0.025, 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)

ntu60_xsub/ntu60_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/ntu60_3danno.pkl', 'xsub_val')
+
+
+"""
+***************
+InfoGCN  v0:
+j jm b bm k km
+2S:     92.96
+4S:     93.27
+6S:     93.51
+***************
+"""
+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], [2, 2, 2, 1, 1, 1])
+print('Top-1', top1(fused, label))
+
+
+"""
+***************
+InfoGCN  v1:
+j j b b k k
+2S:     92.96
+4S:     93.53
+6S:     93.81
+***************
+"""
+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], [1, 1, 1, 1])
+print('Top-1', top1(fused, label))
+
+print('6S')
+fused = comb([j_1, j_2, b_1, b_2, k_1, k_2], [3, 3, 4, 4, 1, 1])
+print('Top-1', top1(fused, label))