def test(epoch):
model.eval()
test_loss = 0
correct = 0
target_arr = []
predict_arr = []
with torch.no_grad():
for batch_idx, (data, target) in enumerate(test_loader):
data, target = data.cuda(), target.cuda()
data, target = Variable(data), Variable(target)
output = model(data)
test_loss += criterion(output, target)
_, pred = output.data.max(1)
batch_correct = pred.eq(target).sum().item()
correct += batch_correct
predict_arr.append(pred.cpu().numpy())
target_arr.append(target.data.cpu().numpy())
writer.add_scalar('/Acc',
100 * float(batch_correct) / data.size(0),
epoch * len(test_loader) + batch_idx)
cm_path = './' + str(epoch) + '_confusematrix'
cm = metrics.get_confuse_matrix(predict_arr, target_arr)
np.save(cm_path, cm)
test_loss /= len(test_loader)
context = 'Test set: Average loss: {:.6f}, Accuracy: {}/{} ({:.2f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
100. * correct / float(len(test_loader.dataset)))
print(context)
log_file.write(context + '\r\n')
def test(epoch):
model.eval()
test_loss, correct = 0, 0
target_arr, predict_arr = [], []
data1 = [0] * args.num_classes
data2 = [0] * args.num_classes
data3 = [0] * args.num_classes
with torch.no_grad():
for i , (data, target) in enumerate(test_loader):
data, target = data.cuda(), target.cuda()
data, target = Variable(data), Variable(target)
output = model(data)
test_loss += criterion(output, target)
_, pred = output.data.max(1)
for i in range(0, target.size(0)):
data1[target[i]] += 1
data3[pred[i]] += 1
if target[i] == pred[i]:
data2[target[i]] += 1
batch_correct = pred.eq(target).sum().item()
correct += batch_correct
predict_arr.append(pred.cpu().numpy())
target_arr.append(target.data.cpu().numpy())
writer.add_scalar('/Acc', 100 * float(batch_correct) / data.size(0), epoch * len(test_loader) + i)
cm_path = args.check_path + '/' + str(epoch) + '_confusematrix'
cm = metrics.get_confuse_matrix(predict_arr, target_arr)
np.save(cm_path, cm)
for j in range(10):
recall = 0 if data1[j] == 0 else data2[j] / data1[j]
precision = 0 if data3[j] == 0 else data2[j] / data3[j]
context = 'Class%1s: recall is %.2f%% (%d in %d), precision is %.2f%% (%d in %d)' % (
str(j), 100 * recall, data2[j], data1[j],
100 * precision, data2[j], data3[j])
print(context)
f.write(context + "\r\n")
test_loss /= len(test_loader)
context = 'Test set: Average loss: {:.6f}, Accuracy: {}/{} ({:.2f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
100. * correct / float(len(test_loader.dataset)))
print(context)
f.write(context + '\r\n')
def validate(epoch, model, clf, args, f, writer, test=True):
model.eval()
transform = transforms.Compose([
transforms.Resize(32),
transforms.ToTensor(),
transforms.Normalize(mean=np.array([0.485, 0.456, 0.406]),
std=np.array([0.229, 0.224, 0.225])),
])
test_set = torchvision.datasets.ImageFolder(
root=args.test_set if test else args.train_set, transform=transform)
test_loader = torch.utils.data.DataLoader(test_set,
batch_size=args.test_batch_size,
shuffle=True)
correct, total = 0, 0
target_arr, predict_arr = [], []
fea, l = torch.zeros(0), torch.zeros(0)
data1 = [0] * args.num_classes
data2 = [0] * args.num_classes
data3 = [0] * args.num_classes
for i, (data, target) in enumerate(test_loader):
data, target = data.cuda(), target.cuda()
data, target = Variable(data), Variable(target)
output = model.forward(data)
predicted = clf.predict(output.data.cpu().numpy())
predict_arr.append(predicted)
target_arr.append(target.data.cpu().numpy())
fea = torch.cat((fea, output.data.cpu()))
l = torch.cat((l, target.data.cpu().float()))
for i in range(0, target.size(0)):
data1[target[i]] += 1
data3[predicted[i]] += 1
if target[i] == predicted[i]:
data2[target[i]] += 1
correct += (torch.tensor(predicted) == target.data.cpu()).sum()
total += target.size(0)
writer.add_scalar('/Acc', 100. * correct / float(total),
epoch * len(test_loader) + i)
k_n = (epoch * 2) if test else (epoch * 2 + 1)
writer.add_embedding(mat=fea, metadata=l, global_step=k_n)
cm_path = args.check_path + '/' + 'train' + str(epoch) + '_confusematrix'
if test:
cm_path = args.check_path + '/' + 'test' + str(
epoch) + '_confusematrix'
cm = metrics.get_confuse_matrix(predict_arr, target_arr)
np.save(cm_path, cm)
for j in range(args.num_classes):
recall = 0 if data1[j] == 0 else data2[j] / data1[j]
precision = 0 if data3[j] == 0 else data2[j] / data3[j]
context = 'Class%1s: recall is %.2f%% (%d in %d), precision is %.2f%% (%d in %d)' % (
str(j), 100 * recall, data2[j], data1[j], 100 * precision,
data2[j], data3[j])
print(context)
f.write(context + "\r\n")
context = 'Accuracy of model in ' + ('test' if test else 'train') + \
' set is {}/{}({:.2f}%)'.format(correct, total, 100. * float( correct) / float(total))
f.write(context + '\r\n')
print(context)