Upload 122 files

finegym/b_1/b_1.py

@@ -0,0 +1,113 @@
+modality = 'b'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/b_1'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/b_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1667686a27b2ef3a7ac7d2ce0505853620eb489ad558a9c32820b2dc465f2e78
+size 5254933

finegym/b_1/best_top1_acc_epoch_138.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1754379abe875e2d57bc6466ceda3a254a57f4e88260bd42c1691b520dcc6015
+size 31999601

finegym/b_2/b_2.py

@@ -0,0 +1,113 @@
+modality = 'b'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/b_2'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/b_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd4b2d4a6b43b9e9c1cbe5d5db7b44799f971661080320e7c22858002bc99093
+size 5254226

finegym/b_2/best_top1_acc_epoch_139.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f51a18f1d5646ec55ef9127d25424ab95aa47280c06a7b0323ddd25a679a3ff
+size 31999601

finegym/b_3/b_3.py

@@ -0,0 +1,113 @@
+modality = 'b'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/b_3'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['b']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/b_3/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbca7f85d66fa3978c79158c363760b4c13af9d2e0d01a5d21fa87704bfdd38b
+size 5253922

finegym/b_3/best_top1_acc_epoch_148.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce2ea406d71cf80ebf90b44e6b69876854ddefc870b0ee56bdc128eaff1303fc
+size 31999601

finegym/bm/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1d334b3d197936a0bdf6324091c167c42a4ef18795428df4b50b17b258ce90e
+size 5257893

finegym/bm/best_top1_acc_epoch_148.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85f8fa42d235cb630ac931492b8a6a08835fea0cd97221f3df638526806f91e3
+size 31999601

finegym/bm/bm.py

@@ -0,0 +1,113 @@
+modality = 'bm'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/bm'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['bm']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['bm']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['bm']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['bm']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['bm']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['bm']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/finegym_ensemble.py

@@ -0,0 +1,68 @@
+from mmcv import load
+import sys
+# Note: please adjust the relative path according to the actual situation.
+sys.path.append('../..')
+from aclnet.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')
+jm = load('jm/best_pred.pkl')
+bm = load('bm/best_pred.pkl')
+km = load('km/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')
+j_3 = load('j_3/best_pred.pkl')
+b_3 = load('b_3/best_pred.pkl')
+k_3 = load('k_3/best_pred.pkl')
+label = load_label('/data/finegym/gym_hrnet.pkl', 'val')
+
+
+"""
+***************
+InfoGCN  v0:
+j jm b bm k km
+2S:     95.48
+4S:     95.80
+6S:     96.01
+***************
+"""
+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], [9, 9, 5, 5])
+print('Top-1', top1(fused, label))
+
+print('6S')
+fused = comb([j_1, b_1, k_1, jm, bm, km], [9, 9, 9, 5, 5, 5])
+print('Top-1', top1(fused, label))
+
+
+"""
+***************
+HD-GCN  v1:
+j b j b j b
+2S:     95.48
+4S:     95.79
+6S:     95.93
+***************
+"""
+print('HD-GCN  v1:')
+print('j b j b j b')
+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, 4, 3, 3])
+print('Top-1', top1(fused, label))
+
+print('6S')
+fused = comb([j_1, b_1, j_2, b_2, j_3, b_3], [4, 4, 3, 3, 8, 8])
+print('Top-1', top1(fused, label))

finegym/j_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb1cca8720d20096212d90cab32335f99ac86240e99869f7c685395e485ccdc5
+size 5256570

finegym/j_1/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a991fcc19307ac0de3c0b7784ad8260dfaa7db5843114bc0673db17aad5b6e5b
+size 16118201

finegym/j_1/j_1.py

@@ -0,0 +1,113 @@
+modality = 'j'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/j_1'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/j_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd20197d42a554c12bc0af8293b23f4a96c30bf6ca028d911b11074c11352432
+size 5257662

finegym/j_2/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bad9ebbd0bcf82a0c88cbab4ab8c28418e150df63fa0be4db2d67e03725efe35
+size 31999601

finegym/j_2/j_2.py

@@ -0,0 +1,113 @@
+modality = 'j'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/j_2'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/j_3/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b7f5f79a0963b9578a46e210f6d17c92b75b738349250975086528b5b7f987a
+size 5255370

finegym/j_3/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b6490fedb1736242b388dbce225ae00b1d62fe536f276a106e4baa1472103fd
+size 16118201

finegym/j_3/j_3.py

@@ -0,0 +1,113 @@
+modality = 'j'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/j_3'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['j']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/jm/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2e4dd79eca7185518373b4ee6b273b476f724304ec175ce47a725dc6f3d92d2
+size 5255302

finegym/jm/best_top1_acc_epoch_147.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:151266b00eca2713ffb4366fcc1348a209ec738956a8390e3fd416af288b2383
+size 31999601

finegym/jm/jm.py

@@ -0,0 +1,113 @@
+modality = 'jm'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/jm'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['jm']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['jm']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['jm']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['jm']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['jm']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['jm']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/k_1/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c06926547b6652db4986b1a7d59f41d80ae7834c149eb0689a0317f5f4e848a
+size 5257632

finegym/k_1/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:487f7f480cc3b6c91ab42dfbef04731c5dfc3c9628e1af31eb58e87f5aca765c
+size 31999601

finegym/k_1/k_1.py

@@ -0,0 +1,113 @@
+modality = 'k'
+graph = 'coco_new'
+work_dir = './work_dirs/test_aclnet/finegym/k_1'
+model = dict(
+    type='RecognizerGCN',
+    backbone=dict(
+        type='GCN_Module',
+        gcn_ratio=0.125,
+        gcn_ctr='T',
+        gcn_ada='T',
+        tcn_ms_cfg=[(3, 1), (3, 2), (3, 3), (3, 4), ('max', 3), '1x1'],
+        graph_cfg=dict(
+            layout='coco_new',
+            mode='random',
+            num_filter=8,
+            init_off=0.04,
+            init_std=0.02)),
+    cls_head=dict(type='SimpleHead', data_cfg='finegym', num_classes=99, in_channels=384))
+dataset_type = 'PoseDataset'
+ann_file = '/data/lhd/pyskl_data/gym/gym_hrnet.pkl'
+left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
+right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
+train_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100),
+    dict(type='PoseDecode'),
+    dict(
+        type='Flip',
+        flip_ratio=0.5,
+        left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+        right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['k']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+val_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['k']),
+    dict(type='FormatGCNInput', num_person=2),
+    dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+    dict(type='ToTensor', keys=['keypoint'])
+]
+test_pipeline = [
+    dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+    dict(type='PoseDecode'),
+    dict(type='Kinetics_Transform'),
+    dict(type='GenSkeFeat', dataset='coco_new', feats=['k']),
+    dict(type='FormatGCNInput', num_person=2),
+    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/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100),
+            dict(type='PoseDecode'),
+            dict(
+                type='Flip',
+                flip_ratio=0.5,
+                left_kp=[1, 3, 5, 7, 9, 11, 13, 15],
+                right_kp=[2, 4, 6, 8, 10, 12, 14, 16]),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['k']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='train'),
+    val=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=1),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['k']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='val'),
+    test=dict(
+        type='PoseDataset',
+        ann_file='/data/lhd/pyskl_data/gym/gym_hrnet.pkl',
+        pipeline=[
+            dict(type='UniformSampleFrames', clip_len=100, num_clips=10),
+            dict(type='PoseDecode'),
+            dict(type='Kinetics_Transform'),
+            dict(type='GenSkeFeat', dataset='coco_new', feats=['k']),
+            dict(type='FormatGCNInput', num_person=2),
+            dict(type='Collect', keys=['keypoint', 'label'], meta_keys=[]),
+            dict(type='ToTensor', keys=['keypoint'])
+        ],
+        split='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', 'mean_class_accuracy'], topk=(1, 5))
+log_config = dict(interval=100, hooks=[dict(type='TextLoggerHook')])
+dist_params = dict(backend='nccl')
+gpu_ids = range(0, 1)

finegym/k_2/best_pred.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5752510808b3d0082f81c86c622811a49bcc15d463858c6f764052114cf58ad2
+size 5256300

finegym/k_2/best_top1_acc_epoch_150.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6560b85bcb87466c359d8a46b7f652c8063378ca94483039eb4bbeed21c0584
+size 31999601