Skip to content

Don't Merge: Compare master and jinshi/PoC #5

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jinzishuai wants to merge 5 commits into
base: master
Choose a base branch
from
Open

Don't Merge: Compare master and jinshi/PoC #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
47c7166
transf_rotations = False to have deterministic results
jinzishuai Oct 18, 2020
eb4028e
running on 2009-5 training data
jinzishuai Oct 19, 2020
cc5989a
Switch to use built-in dataloader since my own does not work
jinzishuai Oct 20, 2020
c24aeb4
restore random rotation
jinzishuai Oct 24, 2020
2d1eb01
adjust test to run zian's model: zerobox#221
jinzishuai Jan 29, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 5 additions & 6 deletions config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,16 +10,15 @@

# data settings
dataset_path = "zerobox_dataset"
class_name = "zerobox_class"
modelname = "zerobox_test"
class_name = "zerobox-2009-5"
modelname = "zerobox-2009-5"

# img_size = (448, 448)
img_size = (480, 270)
img_size = (448, 448)
img_dims = [3] + list(img_size)
add_img_noise = 0.01

# transformation settings
transf_rotations = False
transf_rotations = True
transf_brightness = 0.0
transf_contrast = 0.0
transf_saturation = 0.0
Expand All @@ -36,7 +35,7 @@

# dataloader parameters
n_transforms = 1 # number of transformations per sample in training
n_transforms_test = 64 # number of transformations per sample in testing
n_transforms_test = 16 # number of transformations per sample in testing
batch_size = 1 # actual batch size is this value multiplied by n_transforms(_test)
batch_size_test = 1 # batch_size * n_transforms // n_transforms_test

Expand Down
34 changes: 24 additions & 10 deletions runTest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,8 +16,9 @@
from utils import *
from localization import export_gradient_maps
from torch.autograd import Variable
from sklearn.metrics import roc_curve

def test(model, test_loader):
def test(model, test_loader, target_threshold):
print("Running test")
optimizer = torch.optim.Adam(model.nf.parameters(), lr=c.lr_init, betas=(0.8, 0.8), eps=1e-04, weight_decay=1e-5)
# score_obs = Score_Observer('AUROC')
Expand Down Expand Up @@ -56,10 +57,22 @@ def test(model, test_loader):
print(f"test_labels={test_labels}, is_anomaly={is_anomaly},anomaly_score={anomaly_score}")
# score_obs.update(roc_auc_score(is_anomaly, anomaly_score), epoch,
# print_score=c.verbose or epoch == c.meta_epochs - 1)
# Code to calculate the accurarcy from Lihang's code
is_anomaly_detected = np.array([0 if l < target_threshold else 1 for l in anomaly_score])

if c.grad_map_viz:
print("saving gradient maps...")
export_gradient_maps(model, test_loader, optimizer, -1)
# calculate test accuracy
error_count = 0
for i in range(len(is_anomaly)):
if is_anomaly[i] != is_anomaly_detected[i]:
error_count += 1

test_accuracy = 1 - float(error_count) / len(is_anomaly)

print(f"n_transforms_test = {c.n_transforms_test}, target_threshold={target_threshold}, test_accuracy={test_accuracy}")

# if c.grad_map_viz:
# print("saving gradient maps...")
# export_gradient_maps(model, test_loader, optimizer, -1)

def load_testloader(data_dir_test):
def target_transform(target):
Expand All @@ -80,8 +93,8 @@ def target_transform(target):
class_idx += 1

augmentative_transforms = []
# if c.transf_rotations:
# augmentative_transforms += [transforms.RandomRotation(180)]
if c.transf_rotations:
augmentative_transforms += [transforms.RandomRotation(180)]
if c.transf_brightness > 0.0 or c.transf_contrast > 0.0 or c.transf_saturation > 0.0:
augmentative_transforms += [transforms.ColorJitter(brightness=c.transf_brightness, contrast=c.transf_contrast,
saturation=c.transf_saturation)]
Expand All @@ -100,14 +113,15 @@ def target_transform(target):
# train_set, test_set = load_datasets(c.dataset_path, c.class_name)
# _, test_loader = make_dataloaders(train_set, test_set)

test_loader = load_testloader("group15B.avi/")
# model = torch.load("../zerobox-v2/zerobox_differnet_model.pt", map_location=torch.device('cpu'))
model = torch.load("models/zerobox_test.pt", map_location=torch.device('cpu'))
test_loader = load_testloader("dataset/group15B.avi.Products/test")
model = torch.load("models/zerobox-2010-1-black-yolo_0_0.10_0.05_0.10_0.05_0.9980.pth", map_location=torch.device('cpu'))
# model = torch.load("models/zerobox_test.pt", map_location=torch.device('cpu'))
target_threshold= 3.5129451751708984

print("starting to run tests after loaded model and test dataset")
time_start = time.time()
# model = load_model(c.modelname)
test(model, test_loader)
test(model, test_loader, target_threshold)
time_end = time.time()
time_c = time_end - time_start # 运行所花时间
print("time cost: {:f} s".format(time_c))
Expand Down