Hello, I’m trying to make a predict script based on the test one. But I keep getting error.
This is the Code for the test:
Engine:
def test(self,val_loader, model, criterion):
top1 = AverageMeter()
top5 = AverageMeter()
print_freq = 100
# switch to evaluate mode
model.eval()
with torch.no_grad():
for i, (input, target) in enumerate(val_loader):
target = target.cuda()
input = input.cuda()
# forward
attention_maps, _, output1 = model(input)
refined_input = mask2bbox(attention_maps, input)
_, _, output2 = model(refined_input)
output = (F.softmax(output1, dim=-1)+F.softmax(output2, dim=-1))/2
# measure accuracy and record loss
prec1, prec5 = accuracy(output, target, topk=(1, 5))
loss = criterion(output, target)
top1.update(prec1[0], input.size(0))
top5.update(prec5[0], input.size(0))
if i % print_freq == 0:
print('Test: [{0}/{1}]\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
'Prec@5 {top5.val:.3f} ({top5.avg:.3f})\t'
'Loss {loss.val:.3f} ({loss.avg:.3f})'.format(
i, len(val_loader),
top1=top1, top5=top5,loss=loss))
print(' * Prec@1 {top1.avg:.3f} Prec@5 {top5.avg:.3f} Loss {loss.val:.3f}'
.format(top1=top1, top5=top5,loss=loss))
return top1.avg, top5.avg
def test():
##
engine = Engine()
config = getConfig()
data_config = getDatasetConfig(config.dataset)
# define dataset
transform_test = transforms.Compose([
transforms.Resize((config.image_size, config.image_size)),
transforms.CenterCrop(config.input_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])
val_dataset = CustomDataset(
data_config['val'], data_config['val_root'], transform=transform_test)
val_loader = DataLoader(
val_dataset, batch_size=config.batch_size, shuffle=False, num_workers=config.workers, pin_memory=True)
# define model
if config.model_name == 'inception':
net = inception_v3_bap(pretrained=True, aux_logits=False)
elif config.model_name == 'resnet50':
net = resnet50(pretrained=True)
in_features = net.fc_new.in_features
new_linear = torch.nn.Linear(
in_features=in_features, out_features=val_dataset.num_classes)
net.fc_new = new_linear
# load checkpoint
use_gpu = torch.cuda.is_available() and config.use_gpu
if use_gpu:
net = net.cuda()
gpu_ids = [int(r) for r in config.gpu_ids.split(',')]
if use_gpu and len(gpu_ids) > 1:
net = torch.nn.DataParallel(net, device_ids=gpu_ids)
net = torch.nn.DataParallel(net)
#checkpoint_path = os.path.join(config.checkpoint_path,'model_best.pth.tar')
net.load_state_dict(torch.load(config.checkpoint_path)['state_dict'])
# define loss
# define loss
criterion = torch.nn.CrossEntropyLoss()
if use_gpu:
criterion = criterion.cuda()
prec1, prec5 = engine.test(val_loader, net, criterion)
my codes for predict:
Engine
def predict(self,val_loader, model, criterion):
top1 = AverageMeter()
top5 = AverageMeter()
print_freq = 100
# switch to evaluate mode
model.eval()
with torch.no_grad():
image = Image.open("/content/drive/My Drive/volkswagen-beetle-2017-1280-01-1536873547.jpg").convert('RGB')
rgb_mean = np.mean(image, axis=(0, 1))
rgb_std = np.std(image, axis=(0, 1))
# forward
attention_maps, _, output1 = model(image)
refined_input = mask2bbox(attention_maps, image)
_, _, output2 = model(refined_input)
output = (F.softmax(output1, dim=-1)+F.softmax(output2, dim=-1))/2
# measure accuracy and record loss
prec1, prec5 = accuracy(output, target, topk=(1, 5))
loss = criterion(output, target)
top1.update(prec1[0], input.size(0))
top5.update(prec5[0], input.size(0))
if i % print_freq == 0:
print('Test: [{0}/{1}]\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
'Prec@5 {top5.val:.3f} ({top5.avg:.3f})\t'
'Loss {loss.val:.3f} ({loss.avg:.3f})'.format(
i, len(val_loader),
top1=top1, top5=top5,loss=loss))
return output
def predict():
##
engine = Engine()
config = getConfig()
data_config = getDatasetConfig(config.dataset)
# define dataset
transform_test = transforms.Compose([
transforms.Resize((config.image_size, config.image_size)),
transforms.CenterCrop(config.input_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])
val_dataset = CustomDataset(
data_config['val'], data_config['val_root'], transform=transform_test)
val_loader = DataLoader(
val_dataset, batch_size=config.batch_size, shuffle=False, num_workers=config.workers, pin_memory=True)
# define model
if config.model_name == 'inception':
net = inception_v3_bap(pretrained=True, aux_logits=False)
elif config.model_name == 'resnet50':
net = resnet50(pretrained=True)
in_features = net.fc_new.in_features
new_linear = torch.nn.Linear(
in_features=in_features, out_features=val_dataset.num_classes)
net.fc_new = new_linear
# load checkpoint
use_gpu = torch.cuda.is_available() and config.use_gpu
if use_gpu:
net = net.cuda()
gpu_ids = [int(r) for r in config.gpu_ids.split(',')]
if use_gpu and len(gpu_ids) > 1:
net = torch.nn.DataParallel(net, device_ids=gpu_ids)
#checkpoint_path = os.path.join(config.checkpoint_path,'model_best.pth.tar')
net = torch.nn.DataParallel(net)
net.load_state_dict(torch.load(config.checkpoint_path)['state_dict'])
# define loss
# define loss
criterion = torch.nn.CrossEntropyLoss()
if use_gpu:
criterion = criterion.cuda()
output = engine.predict(val_loader, net, criterion)
arr = numpy.array(lst)
idx = (-arr).argsort()[:5]
print(idx)
I got this error:
predict()
File "train_bap.py", line 251, in predict
output = engine.predict(val_loader, net, criterion)
File "/content/drive/My Drive/WS_DAN_PyTorch-master/utils/engine.py", line 200, in predict
attention_maps, _, output1 = model(image)
File "/usr/local/lib/python3.6/site-packages/torch/nn/modules/module.py", line 550, in __call__
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/site-packages/torch/nn/parallel/data_parallel.py", line 153, in forward
return self.module(*inputs[0], **kwargs[0])
File "/usr/local/lib/python3.6/site-packages/torch/nn/modules/module.py", line 550, in __call__
result = self.forward(*input, **kwargs)
File "/content/drive/My Drive/WS_DAN_PyTorch-master/model/inception_bap.py", line 152, in forward
x_ch0 = torch.unsqueeze(x[:, 0], 1) * (0.229 / 0.5) + (0.485 - 0.5) / 0.5
TypeError: 'Image' object is not subscriptable
