def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=0,
path='../bold_data/BOLD_ijcv/BOLD_public/frames/',
ucf_list='../bold_data/BOLD_ijcv/BOLD_public/annotations/',
ucf_split='04')
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
model.run()
def main():
global arg
arg = parser.parse_args()
print arg
#Preparing DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=12,
# path='/home/ubuntu/data/UCF101/spatial_no_sampled/'
path=
'/media/semanticslab11/hdd1/data/two-stream-action/data/backup/ucf101/', #path to your ucf 101 data
# ucf_list ='/home/ubuntu/lab/pytorch/ucf101_two_stream/github/UCF_list/',
ucf_list='UCF_list/', #here use ucf_list path
ucf_split='01',
)
train_loader, test_loader, test_video = data_loader.run()
#initiating our model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video,
pretrain_finetune=arg.fine_tune_flag,
pretrain_last_layer=arg.last_layer_flag)
#Calling run function to run the model
model.run()
def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=1,
num_workers=1,
path=
'/home/yifu/Documents/Mycode/python/two-stream-pytorch/datasets/jpegs_256/',
ucf_list=
'/home/yifu/Documents/Mycode/python/two-stream-pytorch/twostream/UCF_list/',
ucf_split='01')
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
model.run()
def main(ds_path, trainfile, testfile, prefix):
global arg
arg = parser.parse_args()
print arg
method = EXPERIMENTS.MAIN_AUTHOR
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(BATCH_SIZE=arg.batch_size,
num_workers=12,
path=ds_path,
trainfile=trainfile,
testfile=testfile,
step_size=arg.step,
experiment=method)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video,
num_classes=data_loader.max_cls_ind,
prefix=prefix)
#Training
model.run()
def main():
global arg
arg = parser.parse_args()
print arg
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=8,
path='/home/ubuntu/data/UCF101/spatial_no_sampled/',
ucf_list ='/home/ubuntu/cvlab/pytorch/ucf101_two_stream/github/UCF_list/',
ucf_split ='01',
)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(
nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video
)
#Training
model.run()
def main():
global arg
arg = parser.parse_args()
print arg
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=8,
path='/home/lenovo/xuzeshan/data/UCF101/',
ucf_list ='/home/lenovo/xuzeshan/two-stream-action-recognition/UCF_lis/',
ucf_split ='01',
)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(
nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video
)
#Training
model.run()
def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=12,
path='/media/lsc/DATA/Data/UCF101/spatial_no_sampled/',
ucf_list=
'/media/lsc/DATA/github/two-stream-action-recognition/UCF_list/',
ucf_split='01',
)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
model.run()
def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=8,
path='/media/dataDisk/THUMOS14/THUMOS14_video/thumos14_preprocess/',
train_list=
'./thumos14_list/new_thumos_14_20_one_label_temporal_val.txt',
test_list='./thumos14_list/new_thumos_14_20_one_label_temporal_test.txt'
)
train_loader, test_loader, test_video = data_loader.run()
checkpoint_dir = os.path.join("./record/spatial", "pretrain_ucf101")
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
model.run(checkpoint_dir)
def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=8,
path='/research/dept2/mli/Data/jpegs_256/',
ucf_list='./UCF_list/',
ucf_split='01',
)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
if arg.evaluate:
cudnn.benchmark = True
model.build_model()
model.resume_and_evaluate()
else:
model.run()
def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=0,
path='/home/lb/video_action_recognition/data/jpegs_256',
ucf_list='/home/lb/video_action_recognition/data/datalists',
ucf_split='01',
)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
model.run()
def main():
global arg
arg = parser.parse_args()
print(arg)
#Prepare DataLoader
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=arg.batch_size,
num_workers=0,
path='D:\\Data\\UCF101RGB\\UCF101\\ucf101\\jpegs_256\\',
ucf_list=
'D:\\Radha\\Downloads\\two-stream-action-recognition-master\\UCF_list\\',
ucf_split='01',
)
train_loader, test_loader, test_video = data_loader.run()
#Model
model = Spatial_CNN(nb_epochs=arg.epochs,
lr=arg.lr,
batch_size=arg.batch_size,
resume=arg.resume,
start_epoch=arg.start_epoch,
evaluate=arg.evaluate,
train_loader=train_loader,
test_loader=test_loader,
test_video=test_video)
#Training
model.run()
arg = parser.parse_args()
print(arg)
rgb_preds = arg.rgbPred
opf_preds = arg.optPred
with open(rgb_preds, 'rb') as f:
rgb = pickle.load(f)
f.close()
with open(opf_preds, 'rb') as f:
opf = pickle.load(f)
f.close()
dataloader = dataloader.spatial_dataloader(BATCH_SIZE=1,
num_workers=1,
path=arg.jpeg,
nda_list=arg.NDAlist,
nda_split='01',
crop_size=arg.imgCropSize)
train_loader, val_loader, test_video = dataloader.run()
rgb_video_level_preds = np.zeros((len(rgb.keys()), arg.numClass))
opt_video_level_preds = np.zeros((len(rgb.keys()), arg.numClass))
fuse_video_level_preds = np.zeros((len(rgb.keys()), arg.numClass))
video_level_labels = np.zeros(len(rgb.keys()))
# correct=0
ii = 0
for name in sorted(rgb.keys()):
r = rgb[name]
o = opf[name]
label = int(test_video[name]) - 1
import dataloader
if __name__ == '__main__':
rgb_preds='record/spatial/spatial_video_preds.pickle'
opf_preds = 'record/motion/motion_video_preds.pickle'
with open(rgb_preds,'rb') as f:
rgb =pickle.load(f)
f.close()
with open(opf_preds,'rb') as f:
opf =pickle.load(f)
f.close()
dataloader = dataloader.spatial_dataloader(BATCH_SIZE=1, num_workers=1,
path='/home/ubuntu/data/UCF101/spatial_no_sampled/',
ucf_list='/home/ubuntu/cvlab/pytorch/ucf101_two_stream/github/UCF_list/',
ucf_split='01')
train_loader,val_loader,test_video = dataloader.run()
video_level_preds = np.zeros((len(rgb.keys()),101))
video_level_labels = np.zeros(len(rgb.keys()))
correct=0
ii=0
for name in sorted(rgb.keys()):
r = rgb[name]
o = opf[name]
label = int(test_video[name])-1
video_level_preds[ii,:] = (r+o)
video_level_labels[ii] = label
if __name__ == '__main__':
rgb_preds = 'record/spatial/spatial_video_preds.pickle'
opf_preds = 'record/motion/motion_video_preds.pickle'
with open(rgb_preds, 'rb') as f:
rgb = pickle.load(f)
f.close()
with open(opf_preds, 'rb') as f:
opf = pickle.load(f)
f.close()
dataloader = dataloader.spatial_dataloader(
BATCH_SIZE=1,
num_workers=1,
path='/home/lb/video_action_recognition/data/jpegs_256',
ucf_list='/home/lb/video_action_recognition/data/datalists',
ucf_split='01')
train_loader, val_loader, test_video = dataloader.run()
video_level_preds = np.zeros((len(list(rgb.keys())), 101))
video_level_labels = np.zeros(len(list(rgb.keys())))
correct = 0
ii = 0
final_result = []
for name in sorted(rgb.keys()):
r = rgb[name]
o = opf[name]
from my_network import *
from utils import *
import torch.optim as optim
import dataloader
import torch
import os
from ResidualNetwork import *
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
# Prepare DataLoader
print '*************************Prepare DataLoader********************************'
data_loader = dataloader.spatial_dataloader(
BATCH_SIZE=10,
num_workers=8,
path='/media/ming/DATADRIVE1/Datasets/UCF101 Dataset/jpegs_256/',
UCF_list='/media/ming/DATADRIVE1/UCF101 Multi-stream Code/UCF_list/',
)
train_loader, test_loader, test_video = data_loader.run()
# build the model
model = resnet34(pretrained=False, channel=3).cuda()
print '------model--------'
print model
criterion = nn.CrossEntropyLoss().cuda()
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.9)
def train_1epoch(epoch):
import torch
from utils import *
import dataloader
rgb_video_pred_collections = 'record/spatial/spatial_video_preds_collections_300.pickle'
saliency_video_pred_collections = 'record/saliency/saliency_video_preds_collections_300.pickle'
with open(rgb_video_pred_collections, 'rb') as f:
rgbs = pickle.load(f)
f.close()
with open(saliency_video_pred_collections, 'rb') as f:
saliencys = pickle.load(f)
f.close()
dataloader = dataloader.spatial_dataloader(BATCH_SIZE=1, num_workers=1,
path='/media/ming/DATADRIVE1/KTH Dataset 600/KTH1saliency/',
KTH_list='/media/ming/DATADRIVE1/KTH Multi-stream Code/KTH_list/'
)
train_loader, val_loader, test_video = dataloader.run()
nb_epochs = len(rgbs)
for epoch in range(nb_epochs):
rgb = rgbs[epoch]
saliency = saliencys[epoch]
video_level_preds = np.zeros((len(rgb.keys()), 101))
video_level_labels = np.zeros(len(rgb.keys()))
ii = 0
for name in sorted(rgb.keys()):
if __name__ == '__main__':
rgb_preds='record/spatial/spatial_video_preds.pickle'
#opf_preds = 'record/motion/motion_video_preds.pickle'
opf_preds = 'record/motion_pose/motion_pose_video_preds.pickle'
with open(rgb_preds,'rb') as f:
rgb =pickle.load(f)
f.close()
with open(opf_preds,'rb') as f:
opf =pickle.load(f)
f.close()
dataloader = dataloader.spatial_dataloader(BATCH_SIZE=1, num_workers=1,
path='/home/molly/UCF_data/jpegs_256',
ucf_list='/home/molly/two-stream-action-recognition/UCF_list/',
ucf_split='04')
train_loader,val_loader,test_video = dataloader.run()
video_level_preds = np.zeros((len(rgb.keys()),15))
video_level_labels = np.zeros(len(rgb.keys()))
correct=0
ii=0
for name in sorted(rgb.keys()):
r = rgb[name]
o = opf[name]
label = int(test_video[name])-1
video_level_preds[ii,:] = (r+o)
video_level_labels[ii] = label
rgb_preds='record/spatial/spatial_video_preds.pickle'
opf_preds = 'record/motion/motion_video_preds.pickle'
with open(rgb_preds,'rb') as f:
#rgb =pickle.load(f)
rgb = pickle.load(f,encoding='latin1')
f.close()
with open(opf_preds,'rb') as f:
#opf =pickle.load(f)
opf = pickle.load(f,encoding='latin1')
f.close()
dataloader = dataloader.spatial_dataloader(BATCH_SIZE=1, num_workers=1,
#path='/home/ubuntu/data/UCF101/spatial_no_sampled/',
path='/content/jpegs_256/',
#ucf_list='/home/ubuntu/cvlab/pytorch/ucf101_two_stream/github/UCF_list/',
ucf_list='/content/drive/MyDrive/two_stream/two-stream-action-recognition/UCF_list/',
ucf_split='01')
train_loader,val_loader,test_video = dataloader.run()
video_level_preds = np.zeros((len(rgb.keys()),101))
video_level_labels = np.zeros(len(rgb.keys()))
correct=0
ii=0
for name in sorted(rgb.keys()):
r = rgb[name]
o = opf[name]
label = int(test_video[name])-1
video_level_preds[ii,:] = (r+o)
