def test(filenameSave, model, dataloader_test, args):
for step, (images, labels, filename, filenameGt) in enumerate(dataloader_test):
if (args.cuda):
images = images.cuda()
# labels = labels.cuda()
inputs = Variable(images)
# targets = Variable(labels)
with torch.no_grad():
outputs = model(inputs)
label = outputs[0].max(0)[1].byte().cpu().data
# label_cityscapes = cityscapes_trainIds2labelIds(label.unsqueeze(0))
label_color = Colorize()(label.unsqueeze(0))
# image_transform(label.byte()).save(filenameSave)
# label_save = ToPILImage()(label_color)
label_save = label_color.numpy()
label_save = label_save.transpose(1, 2, 0)
# label_save.save(filenameSave)
images = images.cpu().numpy().squeeze(axis=0).transpose(1, 2, 0)
images = (images*255).astype(np.uint8)
for i in range(len(filename)):
fileSave = '../eval/'+ args.savedir + "/" + filename[i].split("leftImg8bit/")[1]
os.makedirs(os.path.dirname(fileSave), exist_ok=True)
output = cv2.addWeighted(images, 0.4, label_save, 0.6, 0)
cv2.imwrite(fileSave,output)
def inference(model, args):
image_folder = "/media/pandongwei/ExtremeSSD/work_relative/extract_img/2020.10.16_1/"
video_save_path = "/home/pandongwei/work_repository/erfnet_pytorch/eval/"
# parameters about saving video
fourcc = cv2.VideoWriter_fourcc(*'XVID')
out = cv2.VideoWriter(video_save_path + 'output_new.avi', fourcc, 10.0, (640, 480))
cuda = True
model.eval()
paths = []
for root, dirs, files in os.walk(image_folder, topdown=True):
for file in files:
image_path = os.path.join(image_folder, file)
paths.append(image_path)
paths.sort()
font = cv2.FONT_HERSHEY_SIMPLEX
angle_pre = 0
for i, path in enumerate(paths):
start_time = time.time()
image = cv2.imread(path)
image = (image / 255.).astype(np.float32)
image = ToTensor()(image).unsqueeze(0)
if (cuda):
image = image.cuda()
input = Variable(image)
with torch.no_grad():
output = model(input)
label = output[0].max(0)[1].byte().cpu().data
# label_cityscapes = cityscapes_trainIds2labelIds(label.unsqueeze(0))
label_color = Colorize()(label.unsqueeze(0))
label_save = label_color.numpy()
label_save = label_save.transpose(1, 2, 0)
# 加上路径规划
label_save, angle = perception_to_angle(label_save, angle_pre)
# 加一个滤波以防止角度突然跳变 TODO
if abs(angle - angle_pre) > 10:
angle = angle_pre
angle_pre = angle
# label_save.save(filenameSave)
image = image.cpu().numpy().squeeze(axis=0).transpose(1, 2, 0)
image = (image * 255).astype(np.uint8)
output = cv2.addWeighted(image, 0.5, label_save, 0.5, 0)
cv2.putText(output,str(round(angle,3)),(50,50),cv2.FONT_HERSHEY_SIMPLEX,2,(0,0,0),2)
#output = np.hstack([label_save, image])
out.write(output)
print(i, " time: %.2f s" % (time.time() - start_time))
out.release()
def inference(model, image, angle_pre):
# pre-process
image = (image / 255.).astype(np.float32)
image = ToTensor()(image).unsqueeze(0)
image = image.cuda()
input = Variable(image)
# inference
with torch.no_grad():
output = model(input)
# post-process
label = output[0].max(0)[1].byte().cpu().data
label_color = Colorize()(label.unsqueeze(0))
label_save = label_color.numpy()
label_save = label_save.transpose(1, 2, 0)
# 加上路径规划
label_save, angle = perception_to_angle(label_save, angle_pre)
return angle
def ensemble_predict(prob_fns, outfile='sample.png',
out_npy_file='sample.npy', out_vis_file='sample_vis.png',
method='averaging', out_shape=(2048, 1024)):
"""Output predict file from two npy files by the given method."""
probs = [np.load(prob_fn) for prob_fn in prob_fns]
if method == 'averaging':
prob = sum(probs) / len(probs)
elif method == 'nms':
prob = np.max(probs, 0)
# -- output npy --
np.save(out_npy_file, prob)
# -- output label-predict --
pred = prob[:N_CLASS].argmax(0)
img = Image.fromarray(np.uint8(pred))
img = img.resize(out_shape, Image.NEAREST)
img.save(outfile)
# -- output vis-predict --
# ToTensor function: ndarray -> Tensor
# * H, W, C -> C, H, W
# * 0, 255 -> 0, 1
# prob.shape == (20, 512, 1024)
prob_for_tensor = np.transpose(prob[:N_CLASS], (1, 2, 0))
prob_tensor = ToTensor()(prob_for_tensor)
pred = prob_tensor.max(0)[1]
pred = pred.view(1, pred.size()[0], pred.size()[1])
# Colorize function: Tensor -> Tensor
vis_tensor = Colorize()(pred)
# Tensor -> ndarray -> image(save)
vis = np.transpose(vis_tensor.numpy(), (1, 2, 0))
vis_img = Image.fromarray(vis, 'RGB')
vis_img = vis_img.resize(out_shape, Image.NEAREST)
vis_img.save(out_vis_file)
def main(args):
modelpath = args.loadDir + args.loadModel
weightspath = args.loadDir + args.loadWeights
print("Loading model: " + modelpath)
print("Loading weights: " + weightspath)
model = Net(NUM_CLASSES)
model = torch.nn.DataParallel(model)
if (not args.cpu):
model = model.cuda()
#model.load_state_dict(torch.load(args.state))
#model.load_state_dict(torch.load(weightspath)) #not working if missing key
def load_my_state_dict(
model, state_dict
): #custom function to load model when not all dict elements
own_state = model.state_dict()
for name, param in state_dict.items():
if name not in own_state:
continue
own_state[name].copy_(param)
return model
model = load_my_state_dict(model, torch.load(weightspath))
print("Model and weights LOADED successfully")
model.eval()
if (not os.path.exists(args.datadir)):
print("Error: datadir could not be loaded")
loader = DataLoader(cityscapes(args.datadir,
input_transform_cityscapes,
target_transform_cityscapes,
subset=args.subset),
num_workers=args.num_workers,
batch_size=args.batch_size,
shuffle=False)
# For visualizer:
# must launch in other window "python3.6 -m visdom.server -port 8097"
# and access localhost:8097 to see it
if (args.visualize):
vis = visdom.Visdom()
for step, (images, labels, filename, filenameGt) in enumerate(loader):
if (not args.cpu):
images = images.cuda()
#labels = labels.cuda()
inputs = Variable(images)
#targets = Variable(labels)
with torch.no_grad():
outputs = model(inputs)
label = outputs[0].max(0)[1].byte().cpu().data
#label_cityscapes = cityscapes_trainIds2labelIds(label.unsqueeze(0))
label_color = Colorize()(label.unsqueeze(0))
filenameSave = "./save_color/" + filename[0].split("leftImg8bit/")[1]
os.makedirs(os.path.dirname(filenameSave), exist_ok=True)
#image_transform(label.byte()).save(filenameSave)
label_save = ToPILImage()(label_color)
label_save.save(filenameSave)
if (args.visualize):
vis.image(label_color.numpy())
print(step, filenameSave)
model.eval()
# 10 13 48 86 101
img = Image.open("./data/VOC2012test/JPEGImages/2008_000101.jpg").convert("RGB")
original_size = img.size
img.save("original.png")
img = img.resize((256, 256), Image.BILINEAR)
img = ToTensor()(img)
img = Variable(img).unsqueeze(0)
outputs = model(img)
# 22 256 256
for i, output in enumerate(outputs):
output = output[0].data.max(0)[1]
output = Colorize()(output)
output = np.transpose(output.numpy(), (1, 2, 0))
img = Image.fromarray(output, "RGB")
if i == 0:
img = img.resize(original_size, Image.NEAREST)
img.save("test-%d.png" % i)
'''
for index, (imgs, name, size) in tqdm(enumerate(testloader)):
imgs = Variable(imgs.cuda())
outputs = model(imgs)
output = outputs[0][0].data.max(0)[1]
output = Colorize()(output)
print(output)
output = np.transpose(output.numpy(), (1, 2, 0))
def main(args):
modelpath = args.loadDir + args.loadModel
weightspath = args.loadDir + args.loadWeights
print("Loading model: " + modelpath)
print("Loading weights: " + weightspath)
#Import ERFNet model from the folder
#Net = importlib.import_module(modelpath.replace("/", "."), "ERFNet")
model = ERFNet(NUM_CLASSES)
model = torch.nn.DataParallel(model)
if (not args.cpu):
model = model.cuda()
#model.load_state_dict(torch.load(args.state))
#model.load_state_dict(torch.load(weightspath)) #not working if missing key
fourcc = cv2.VideoWriter_fourcc(*'MP4V') # Save as video
out = cv2.VideoWriter('output.mp4', fourcc, 20.0, (640, 352))
def load_my_state_dict(
model, state_dict
): #custom function to load model when not all dict elements
own_state = model.state_dict()
for name, param in state_dict.items():
if name not in own_state:
continue
own_state[name].copy_(param)
return model
model = load_my_state_dict(
model, torch.load(weightspath, map_location=torch.device('cpu')))
print("Model and weights LOADED successfully")
model.eval()
# loader = DataLoader(cityscapes(args.datadir, input_transform_cityscapes, target_transform_cityscapes, subset=args.subset),
# num_workers=args.num_workers, batch_size=args.batch_size, shuffle=False)
# For visualizer:
# must launch in other window "python3.6 -m visdom.server -port 8097"
# and access localhost:8097 to see it
# if (args.visualize):
# vis = visdom.Visdom()
# for step, (images, labels, filename, filenameGt) in enumerate(loader):
# cap = cv2.VideoCapture(0)
cap = cv2.VideoCapture('project_video_trimmed.mp4')
while (True):
# Capture frame-by-frame
ret, images = cap.read()
# print(images.shape)
images = trans(images)
images = images.float()
images = images.view((1, 3, 352, 640)) # vidoe
# Our operations on the frame come here
if (not args.cpu):
images = images.cuda()
inputs = Variable(images)
with torch.no_grad():
outputs = model(inputs)
label = outputs[0].max(0)[1].byte().cpu().data
#label_cityscapes = cityscapes_trainIds2labelIds(label.unsqueeze(0))
label_color = Colorize()(label.unsqueeze(0))
frame = label_color.numpy().transpose(1, 2, 0)
# label_save = ToPILImage()(label_color)
# label_save.save("result_1.png")
# Display the resulting frame
cv2.imshow('Segmented Image', frame)
out.write(frame) # To save video file
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
out.release()
cv2.destroyAllWindows()
if (args.visualize):
vis.image(label_color.numpy())
print(step, filenameSave)
def main(args):
modelpath = args.loadDir + args.loadModel
weightspath = args.loadDir + args.loadWeights
if (not os.path.exists(args.datadir)):
print("Error: datadir could not be loaded")
else:
print("Loading model: " + modelpath)
print("Loading weights: " + weightspath)
# Import ERFNET
model = ERFNet(NUM_CLASSES)
model = torch.nn.DataParallel(model)
if (not args.cpu):
model = model.cuda()
def load_my_state_dict(
model, state_dict
): #custom function to load model when not all dict elements
own_state = model.state_dict()
for name, param in state_dict.items():
if name not in own_state:
continue
own_state[name].copy_(param)
return model
model = load_my_state_dict(model, torch.load(weightspath))
print("Model and weights LOADED successfully")
# Set model to Evaluation mode
model.eval()
# Setup the dataset loader
### RELLIS-3D Dataloader
enc = False
loader_test = custom_datasets.setup_loaders(args, enc)
# For visualizer:
# must launch in other window "python3.6 -m visdom.server -port 8097"
# and access localhost:8097 to see it
if (args.visualize):
vis = visdom.Visdom()
for step, (images, labels, img_name, _) in enumerate(loader_test):
start_time = time.time()
if (not args.cpu):
images = images.cuda()
#labels = labels.cuda()
inputs = Variable(images)
with torch.no_grad():
outputs = model(inputs)
label = outputs[0].max(0)[1].byte().cpu().data
label_color = Colorize()(label.unsqueeze(0))
eval_save_path = "./save_colour_rellis/"
if not os.path.exists(eval_save_path):
os.makedirs(eval_save_path)
_, file_name = os.path.split(img_name[0])
file_name = file_name + ".png"
#image_transform(label.byte()).save(filenameSave)
label_save = ToPILImage()(label_color)
label_save.save(os.path.join(eval_save_path, file_name))
if (args.visualize):
vis.image(label_color.numpy())
if step != 0: #first run always takes some time for setup
fwt = time.time() - start_time
time_train.append(fwt)
print("Forward time per img (b=%d): %.3f (Mean: %.3f)" %
(args.batch_size, fwt / args.batch_size,
sum(time_train) / len(time_train) / args.batch_size))
print(step, os.path.join(eval_save_path, file_name))
imgs = Variable(imgs.cuda())
start_time = time.time()
outputs_4, outputs_2, outputs = model(imgs)
end_time = time.time()
if j > 9:
avr_time += (end_time - start_time)
if j == 109:
print avr_time/100.0
exit()
# 22 256 256
filename = list(names)[0]
for i, (output_4, output_2, output) in enumerate(zip(outputs_4, outputs_2, outputs)):
output_4 = output_4.data.max(0)[1]
output_4 = Colorize()(output_4)
output_4 = np.transpose(output_4.numpy(), (1, 2, 0))
img_4 = Image.fromarray(output_4, "RGB")
output_2 = output_2.data.max(0)[1]
output_2 = Colorize()(output_2)
output_2 = np.transpose(output_2.numpy(), (1, 2, 0))
img_2 = Image.fromarray(output_2, "RGB")
output = output.data.max(0)[1]
output = Colorize()(output)
output = np.transpose(output.numpy(), (1, 2, 0))
img = Image.fromarray(output, "RGB")
if i == 0:
img_4 = img_4.resize((width, height), Image.NEAREST)
def main(args):
modelpath = args.loadDir + args.loadModel
#weightspath = args.loadDir + args.loadWeights #TODO
weightspath = "../save/feriburgForest_3/model_best.pth"
print("Loading model: " + modelpath)
print("Loading weights: " + weightspath)
#Import ERFNet model from the folder
#Net = importlib.import_module(modelpath.replace("/", "."), "ERFNet")
model = ERFNet(NUM_CLASSES)
model = torch.nn.DataParallel(model)
if (not args.cpu):
model = model.cuda()
#model.load_state_dict(torch.load(args.state))
#model.load_state_dict(torch.load(weightspath)) #not working if missing key
def load_my_state_dict(
model, state_dict
): #custom function to load model when not all dict elements
own_state = model.state_dict()
for name, param in state_dict.items():
if name not in own_state:
continue
own_state[name].copy_(param)
return model
model = load_my_state_dict(model, torch.load(weightspath))
print("Model and weights LOADED successfully")
model.eval()
if (not os.path.exists(args.datadir)):
print("Error: datadir could not be loaded")
loader = DataLoader(freiburgForest(args.datadir,
input_transform_cityscapes,
target_transform_cityscapes,
subset=args.subset),
num_workers=args.num_workers,
batch_size=args.batch_size,
shuffle=False)
# For visualizer:
# must launch in other window "python3.6 -m visdom.server -port 8097"
# and access localhost:8097 to see it
if (args.visualize):
vis = visdom.Visdom()
for step, (images, labels, filename, filenameGt) in enumerate(loader):
if (not args.cpu):
images = images.cuda()
#labels = labels.cuda()
inputs = Variable(images)
#targets = Variable(labels)
with torch.no_grad():
outputs = model(inputs)
print(outputs.shape)
label = outputs[0].max(0)[1].byte().cpu().data
#label_cityscapes = cityscapes_trainIds2labelIds(label.unsqueeze(0))
label_color = Colorize()(label.unsqueeze(0))
filenameSave = "./freiburgforest_1/" + filename[0].split(
"freiburg_forest_annotated/")[1]
os.makedirs(os.path.dirname(filenameSave), exist_ok=True)
#image_transform(label.byte()).save(filenameSave)
print(label_color.shape)
# label_save = ToPILImage()(label_color)
label_save = label_color.numpy()
label_save = label_save.transpose(1, 2, 0)
print(label_save.shape)
# label_save.save(filenameSave)
images = images.cpu().numpy().squeeze(axis=0)
images = images.transpose(1, 2, 0)
# print(images.shape)
# print(label_save.shape)
plt.figure(figsize=(10.24, 5.12), dpi=100)
plt.imshow(images)
plt.imshow(label_save, alpha=0.6)
plt.axis('off')
# plt.show()
plt.savefig(filenameSave, dpi=100)
plt.close()
if (args.visualize):
vis.image(label_color.numpy())
print(step, filenameSave)
def main(args):
modelpath = args.loadDir + args.loadModel
weightspath = args.loadDir + args.loadWeights
print("Loading model: " + modelpath)
print("Loading weights: " + weightspath)
model = Net(NUM_CLASSES)
model = torch.nn.DataParallel(model)
if (not args.cpu):
model = model.cuda()
# model.load_state_dict(torch.load(args.state))
# model.load_state_dict(torch.load(weightspath)) #not working if missing key
def load_my_state_dict(
model, state_dict
): # custom function to load model when not all dict elements
own_state = model.state_dict()
for name, param in state_dict.items():
if name not in own_state:
continue
own_state[name].copy_(param)
return model
model = load_my_state_dict(model, torch.load(weightspath))
print("Model and weights LOADED successfully")
model.eval()
if (not os.path.exists(args.datadir)):
print("Error: datadir could not be loaded")
# loader = DataLoader(
# cityscapes('/home/liqi/PycharmProjects/LEDNet/4.png', input_transform_cityscapes, target_transform_cityscapes, subset=args.subset),
# num_workers=args.num_workers, batch_size=1 ,shuffle=False)
input_transform_cityscapes = Compose([
Resize((512, 1024), Image.BILINEAR),
ToTensor(),
# Normalize([.485, .456, .406], [.229, .224, .225]),
])
name = "4.png"
with open(image_path_city('/home/gongyiqun/images', name), 'rb') as f:
images = load_image(f).convert('RGB')
images = input_transform_cityscapes(images)
# For visualizer:
# must launch in other window "python3.6 -m visdom.server -port 8097"
# and access localhost:8097 to see it
if (args.visualize):
vis = visdom.Visdom()
if (not args.cpu):
images = images.cuda()
# labels = labels.cuda()
a = torch.unsqueeze(images, 0)
inputs = Variable(a)
# targets = Variable(labels)
with torch.no_grad():
outputs = model(inputs)
label = outputs[0].max(0)[1].byte().cpu().data
# label_cityscapes = cityscapes_trainIds2labelIds(label.unsqueeze(0))
label_color = Colorize()(label.unsqueeze(0))
filenameSave = "./save_color/" + "Others/" + name
os.makedirs(os.path.dirname(filenameSave), exist_ok=True)
# image_transform(label.byte()).save(filenameSave)
label_save = ToPILImage()(label_color)
label_save = label_save.resize((1241, 376), Image.BILINEAR)
# label_save = cv2.resize(label_save, (376, 1224),interpolation=cv2.INTER_AREA)
label_save.save(filenameSave)
if (args.visualize):
vis.image(label_color.numpy())