APBGCN

Attention-Based Parts-oriented Graph Convolution Networks

Dataset

download ntu rgb+d 60 action recognition from skeleton from http://rose1.ntu.edu.sg/datasets/actionRecognition.asp

or use google drive

NTU60 NTU120

uzip data as the following file structure: APBGCN/raw/.*skeleton (create "raw" directory under APBGCN and put skeleton files)

run the code below to generate dataset:

python datagen.py

Training

git fetch and checkout to "distributed" branch

python train_dist.py -#distributed training

Configuration

parser.set_defaults(gpu=True,
                        batch_size=128,
                        dataset_name='NTU',
                        dataset_root=osp.join(os.getcwd()),
                        load_model=False,
                        in_channels=9,
                        num_enc_layers=5,
                        num_conv_layers=2,
                        weight_decay=4e-5,
                        drop_rate=[0.4, 0.4, 0.4, 0.4],  # linear_attention, sparse_attention, add_norm, ffn
                        hid_channels=64,
                        out_channels=64,
                        heads=8,
                        data_parallel=False,
                        cross_k=5,
                        mlp_head_hidden=128)

parser.set_defaults(gpu=True,
                        batch_size=128,
                        dataset_name='NTU',
                        dataset_root=osp.join(os.getcwd()),
                        load_model=False,
                        in_channels=9,
                        num_enc_layers=5,
                        num_conv_layers=2,
                        weight_decay=4e-5,
                        drop_rate=[0.4, 0.4, 0.4, 0.4],  # linear_attention, sparse_attention, add_norm, ffn
                        hid_channels=128,
                        out_channels=128,
                        heads=8,
                        data_parallel=False,
                        cross_k=5,
                        mlp_head_hidden=128)