loss.backward()
optimizer.step()
print('Train batch loss: ', train_batch_loss[-1])
train_loss.append(sum(train_batch_loss) / len(train_batch_loss))
print('***Train loss***: ', train_loss[-1])
val_batch_loss = list()
for X_batch_val, gt_score, gt_geo in val_dataloader:
X_batch_val = X_batch_val.to(device)
gt_score = gt_score.to(device)
gt_geo = gt_geo.to(device)
pred_score, pred_geo = model(X_batch_val)
loss = loss_fn(gt_score, pred_score, gt_geo, pred_geo)
val_batch_loss.append(loss.item())
torch.save(model, './east.pt')
print('Val batch loss: ', val_batch_loss[-1])
val_loss.append(sum(val_batch_loss) / len(val_batch_loss))
print('***Validation loss**: ', val_loss[-1])
if best_val_loss > val_loss[-1]:
best_val_loss = val_loss[-1]
torch.save(model.state_dict(), './east1.pt')
print('Save!')
print('best val loss: ', best_val_loss)
def main():
hmean = .0
is_best = False
warnings.simplefilter('ignore', np.RankWarning)
# Prepare for dataset
print('EAST <==> Prepare <==> DataLoader <==> Begin')
# train_root_path = os.path.abspath(os.path.join('./dataset/', 'train'))
train_root_path = cfg.dataroot
train_img = os.path.join(train_root_path, 'img')
train_gt = os.path.join(train_root_path, 'gt')
trainset = custom_dset(train_img, train_gt)
train_loader = DataLoader(trainset,
batch_size=cfg.train_batch_size_per_gpu *
cfg.gpu,
shuffle=True,
collate_fn=collate_fn,
num_workers=cfg.num_workers)
print('EAST <==> Prepare <==> Batch_size:{} <==> Begin'.format(
cfg.train_batch_size_per_gpu * cfg.gpu))
print('EAST <==> Prepare <==> DataLoader <==> Done')
# test datalodaer
"""
for i in range(100000):
for j, (a,b,c,d) in enumerate(train_loader):
print(i, j,'/',len(train_loader))
"""
# Model
print('EAST <==> Prepare <==> Network <==> Begin')
model = East()
model = nn.DataParallel(model, device_ids=cfg.gpu_ids)
model = model.cuda()
init_weights(model, init_type=cfg.init_type)
cudnn.benchmark = True
criterion = LossFunc()
optimizer = torch.optim.Adam(model.parameters(), lr=cfg.lr)
scheduler = lr_scheduler.StepLR(optimizer, step_size=10000, gamma=0.94)
# init or resume
if cfg.resume and os.path.isfile(cfg.checkpoint):
weightpath = os.path.abspath(cfg.checkpoint)
print(
"EAST <==> Prepare <==> Loading checkpoint '{}' <==> Begin".format(
weightpath))
checkpoint = torch.load(weightpath)
start_epoch = checkpoint['epoch']
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
print(
"EAST <==> Prepare <==> Loading checkpoint '{}' <==> Done".format(
weightpath))
else:
start_epoch = 0
print('EAST <==> Prepare <==> Network <==> Done')
for epoch in range(start_epoch, cfg.max_epochs):
train(train_loader, model, criterion, scheduler, optimizer, epoch)
if epoch % cfg.eval_iteration == 0:
# create res_file and img_with_box
output_txt_dir_path = predict(model, criterion, epoch)
# Zip file
submit_path = MyZip(output_txt_dir_path, epoch)
# submit and compute Hmean
hmean_ = compute_hmean(submit_path)
if hmean_ > hmean:
is_best = True
state = {
'epoch': epoch,
'state_dict': model.state_dict(),
'optimizer': optimizer.state_dict(),
'is_best': is_best,
}
save_checkpoint(state, epoch)
def main():
warnings.simplefilter('ignore', np.RankWarning)
#Model
video_root_path = os.path.abspath('./dataset/train/')
video_name_list = sorted(
[p for p in os.listdir(video_root_path) if p.split('_')[0] == 'Video'])
#print('video_name_list', video_name_list)
print('EAST <==> Prepare <==> Network <==> Begin')
model = East()
AGD_model = AGD()
model = nn.DataParallel(model, device_ids=cfg.gpu_ids)
#AGD_model = nn.DataParallel(AGD_model, device_ids=cfg.gpu_ids)
model = model.cuda()
AGD_model = AGD_model.cuda()
init_weights(model, init_type=cfg.init_type)
cudnn.benchmark = True
criterion1 = LossFunc()
#
criterion2 = Ass_loss()
optimizer1 = torch.optim.Adam(model.parameters(), lr=cfg.lr)
optimizer2 = torch.optim.Adam(AGD_model.parameters(), lr=cfg.lr)
scheduler = lr_scheduler.StepLR(optimizer1, step_size=10000, gamma=0.94)
# init or resume
if cfg.resume and os.path.isfile(cfg.checkpoint):
weightpath = os.path.abspath(cfg.checkpoint)
print(
"EAST <==> Prepare <==> Loading checkpoint '{}' <==> Begin".format(
weightpath))
checkpoint = torch.load(weightpath)
start_epoch = checkpoint['epoch']
model.load_state_dict(checkpoint['state_dict'])
#AGD_model.load_state_dict(checkpoint['model2.state_dict'])
optimizer1.load_state_dict(checkpoint['optimizer'])
#optimizer2.load_state_dict(checkpoint['optimizer2'])
print(
"EAST <==> Prepare <==> Loading checkpoint '{}' <==> Done".format(
weightpath))
else:
start_epoch = 0
print('EAST <==> Prepare <==> Network <==> Done')
for epoch in range(start_epoch + 1, cfg.max_epochs):
for video_name in video_name_list:
print(
'EAST <==> epoch:{} <==> Prepare <==> DataLoader <==>{} Begin'.
format(epoch, video_name))
trainset = custom_dset(os.path.join(video_root_path, video_name))
#sampler = sampler_for_video_clip(len(trainset))
train_loader = DataLoader(trainset,
batch_size=cfg.train_batch_size_per_gpu *
cfg.gpu,
shuffle=False,
collate_fn=collate_fn,
num_workers=cfg.num_workers,
drop_last=True)
print('EAST <==> Prepare <==> Batch_size:{} <==> Begin'.format(
cfg.train_batch_size_per_gpu * cfg.gpu))
print(
'EAST <==> epoch:{} <==> Prepare <==> DataLoader <==>{} Done'.
format(epoch, video_name))
train(train_loader, model, AGD_model, criterion1, criterion2,
scheduler, optimizer1, optimizer2, epoch)
'''
for i, (img, score_map, geo_map, training_mask, coord_ids) in enumerate(train_loader):
print('i{} img.shape:{} geo_map.shape{} training_mask.shape{} coord_ids.len{}'.format(i, score_map.shape, geo_map.shape, training_mask.shape, len(coord_ids)))
'''
if epoch % cfg.eval_iteration == 0:
state = {
'epoch': epoch,
'model1.state_dict': model.state_dict(),
'model2.state_dict': AGD_model.state_dict(),
'optimizer1': optimizer1.state_dict(),
'optimizer2': optimizer2.state_dict()
}
save_checkpoint(state, epoch)