def fast_evaluate(args, basedir, contents, idx=0):
# basedir to save the data
style_model = Net(ngf=args.ngf)
style_model.load_state_dict(torch.load(args.model), False)
style_model.eval()
if args.cuda:
style_model.cuda()
style_loader = StyleLoader(args.style_folder,
args.style_size,
cuda=args.cuda)
for content_image in contents:
idx += 1
content_image = utils.tensor_load_rgbimage(content_image,
size=args.content_size,
keep_asp=True).unsqueeze(0)
if args.cuda:
content_image = content_image.cuda()
content_image = Variable(utils.preprocess_batch(content_image))
for isx in range(style_loader.size()):
style_v = Variable(style_loader.get(isx).data)
style_model.setTarget(style_v)
output = style_model(content_image)
filename = os.path.join(basedir, "{}_{}.png".format(idx, isx + 1))
utils.tensor_save_bgrimage(output.data[0], filename, args.cuda)
print(filename)
def run_demo(args, mirror=False):
style_model = Net(ngf=args.ngf)
style_model.load_state_dict(torch.load(args.model))
style_model.eval()
if args.cuda:
style_loader = StyleLoader(args.style_folder, args.style_size)
style_model.cuda()
else:
style_loader = StyleLoader(args.style_folder, args.style_size, False)
# Define the codec and create VideoWriter object
height = args.demo_size
width = int(4.0 / 3 * args.demo_size)
swidth = int(width / 4)
sheight = int(height / 4)
if args.record:
fourcc = cv2.VideoWriter_fourcc('F', 'M', 'P', '4')
out = cv2.VideoWriter('output.mp4', fourcc, 20.0, (2 * width, height))
cam = cv2.VideoCapture(0)
cam.set(3, width)
cam.set(4, height)
key = 0
idx = 0
while True:
# read frame
idx += 1
ret_val, img = cam.read()
if mirror:
img = cv2.flip(img, 1)
cimg = img.copy()
img = np.array(img).transpose(2, 0, 1)
# changing style
if idx % 20 == 1:
style_v = style_loader.get(int(idx / 20))
style_v = Variable(style_v.data)
style_model.setTarget(style_v)
img = torch.from_numpy(img).unsqueeze(0).float()
if args.cuda:
img = img.cuda()
img = Variable(img)
img = style_model(img)
if not args.cuda:
simg = style_v.cpu().data[0].numpy()
img = img.cpu().clamp(0, 255).data[0].numpy()
else:
simg = style_v.data().numpy()
img = img.clamp(0, 255).data[0].numpy()
img = img.transpose(1, 2, 0).astype('uint8')
simg = simg.transpose(1, 2, 0).astype('uint8')
# display
simg = cv2.resize(simg, (swidth, sheight),
interpolation=cv2.INTER_CUBIC)
cimg[0:sheight, 0:swidth, :] = simg
img = np.concatenate((cimg, img), axis=1)
cv2.imshow('MSG Demo', img)
# cv2.imwrite('stylized/%i.jpg'%idx,img)
key = cv2.waitKey(1)
if args.record:
out.write(img)
# Hit 'q' on the keyboard to quit!
if key & 0xFF == ord('q'):
break
cam.release()
if args.record:
out.release()
cv2.destroyAllWindows()
def run_demo(args, mirror=False):
style_model = Net(ngf=args.ngf)
input('style_model =')
style_model.load_state_dict(torch.load(args.model))
input('style_model.load')
style_model.eval()
if args.cuda:
style_loader = StyleLoader(args.style_folder, args.style_size)
style_model.cuda()
else:
style_loader = StyleLoader(args.style_folder, args.style_size, False)
# Define the codec and create VideoWriter object
height = 480
width = 640
fps = 24
# cam = cv2.VideoCapture(0)
# cam.set(3, width)
# cam.set(4, height)
video = cv2.VideoCapture('tvMesh2.mp4')
video.set(3, width)
video.set(4, height)
video.set(5, fps)
# print(video.get(5))
if args.record:
fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v')
out = cv2.VideoWriter('breakWhiteConv.mov', fourcc, 24, (640, 480),
True)
key = 0
idx = 0
counter = 1
while True:
# read frame
idx += 1
ret_val, frame = video.read()
# ret_val, frame = cam.read()
counter += 1
if counter == video.get(7):
video.set(1, 0)
counter = 0
img = np.array(frame).transpose(2, 0, 1)
# changing style
if idx % 20 == 1:
style_v = style_loader.get(int(idx / 20))
style_v = Variable(style_v.data)
style_model.setTarget(style_v)
img = torch.from_numpy(img).unsqueeze(0).float()
if args.cuda:
img = img.cuda()
img = Variable(img)
img = style_model(img)
if args.cuda:
# simg = style_v.cpu().data[0].numpy()
img = img.cpu().clamp(0, 255).data[0].numpy()
else:
# simg = style_v.data().numpy()
img = img.clamp(0, 255).data[0].numpy()
img = img.transpose(1, 2, 0).astype('uint8')
# display
# simg = cv2.resize(simg,(swidth, sheight), interpolation = cv2.INTER_CUBIC)
# cimg[0:sheight,0:swidth,:]=simg
# fimg = np.concatenate((cimg,img),axis=1)
if args.record:
out.write(img)
cv2.imshow('MSG Demo', img)
#cv2.imwrite('stylized/%i.jpg'%idx,img)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cam.release()
if args.record:
out.release()
cv2.destroyAllWindows()
def run_demo(args, mirror=False):
# style_model = Net(ngf=args.ngf)
# style_model.load_state_dict(torch.load(args.model))
model_dict = torch.load(args.model)
model_dict_clone = model_dict.copy() # We can't mutate while iterating
for key, value in model_dict_clone.items():
if key.endswith(('running_mean', 'running_var')):
del model_dict[key]
style_model = Net(ngf=args.ngf)
style_model.load_state_dict(model_dict, False)
style_model.eval()
if args.cuda:
style_loader = StyleLoader(args.style_folder, args.style_size)
style_model.cuda()
else:
style_loader = StyleLoader(args.style_folder, args.style_size, False)
# Define the codec and create VideoWriter object
height = args.demo_size
width = int(4.0 / 3 * args.demo_size)
swidth = int(width / 4)
sheight = int(height / 4)
if args.record:
fourcc = cv2.VideoWriter_fourcc('F', 'M', 'P', '4')
out = cv2.VideoWriter('output.mp4', fourcc, 20.0, (2 * width, height))
cam = cv2.VideoCapture(0)
cam.set(3, width)
cam.set(4, height)
key = 0
idx = 0
while True:
# read frame
idx += 1
ret_val, img = cam.read()
if mirror:
img = cv2.flip(img, 1)
cimg = img.copy()
img = np.array(img).transpose(2, 0, 1)
# changing style
if idx % 10 == 1:
if int(idx / 10) == 21:
break
style_v = style_loader.get(int(idx / 10))
print('style version ----------------> ', idx)
style_v = Variable(style_v.data)
# print('style change to ', style_v)
style_model.setTarget(style_v)
img = torch.from_numpy(img).unsqueeze(0).float()
if args.cuda:
img = img.cuda()
img = Variable(img)
## handle image
img = style_model(img)
if args.cuda:
simg = style_v.cpu().data[0].numpy()
img = img.cpu().clamp(0, 255).data[0].numpy()
else:
simg = style_v.data().numpy()
img = img.clamp(0, 255).data[0].numpy()
img = img.transpose(1, 2, 0).astype('uint8')
simg = simg.transpose(1, 2, 0).astype('uint8')
# display
simg = cv2.resize(simg, (swidth, sheight),
interpolation=cv2.INTER_CUBIC)
cimg[0:sheight, 0:swidth, :] = simg
img = np.concatenate((cimg, img), axis=1)
cv2.imshow('MSG Demo', img)
#cv2.imwrite('stylized/%i.jpg'%idx,img)
key = cv2.waitKey(1)
if args.record:
out.write(img)
if key == 27:
break
cam.release()
if args.record:
out.release()
cv2.destroyAllWindows()
def run_demo(cuda, record):
if cuda:
ctx = mx.gpu(0); os.environ['MXNET_CUDNN_AUTOTUNE_DEFAULT'] = '0'
else:
ctx = mx.cpu(0)
model = 'models/21styles.params'
ngf = 128
style_size = 512
style_folder = 'images/styles/'
demo_size = 480 # demo window height
mirror = False
style_loader = StyleLoader(style_folder, style_size, ctx)
style_model = Net(ngf=ngf)
style_model.load_parameters(model, ctx=ctx)
# Define the codec and create VideoWriter object
height = demo_size
width = int(4.0/3*demo_size)
swidth = int(width/4); sheight = int(height/4)
if record:
vDir, vFile = './video/', 'output21.mp4'
vPath = vDir + vFile
try:
if not os.path.exists(vDir):
os.makedirs(vDir)
except OSError as e: print(e); sys.exit(1)
fourcc = cv2.VideoWriter_fourcc('m','p','4','v')
out = cv2.VideoWriter(vPath, fourcc, 20.0, (2*width, height))
cam = cv2.VideoCapture(0)
cam.set(3, width); cam.set(4, height)
key, idx = 0, 0
wName = 'STYLIZED VIDEO W:'+ str(width) +' H:'+ str(height)
cv2.namedWindow(wName,cv2.WINDOW_NORMAL)
cv2.resizeWindow(wName, width*2, height)
while True:
# read frame
idx += 1
ret_val, img = cam.read()
if mirror:
img = cv2.flip(img, 1)
cimg = img.copy()
img = np.array(img).transpose(2, 0, 1).astype(float)
img = F.expand_dims(mx.nd.array(img, ctx=ctx), 0)
# changing styles
if idx%20 == 1:
style_v = style_loader.get(int(idx/20))
style_model.set_target(style_v)
img = style_model(img)
simg = np.squeeze(style_v.asnumpy())
simg = simg.transpose(1, 2, 0).astype('uint8')
img = F.clip(img[0], 0, 255).asnumpy()
img = img.transpose(1, 2, 0).astype('uint8')
# display
simg = cv2.resize(simg,(swidth, sheight), interpolation = cv2.INTER_CUBIC)
simg = cv2.cvtColor(simg, cv2.COLOR_BGR2RGB)
cimg[0:sheight,0:swidth,:]=simg
img = np.concatenate((cimg, cv2.COLOR_BGR2RGB)),axis=1)
cv2.imshow(wName, img)
key = cv2.waitKey(1)
if record:
out.write(img)
if key == 27: # Esc
break
def run_demo():
style_model = Net(ngf=this_ngf)
model_dict = torch.load(model21)
model_dict_clone = model_dict.copy()
for key, value in model_dict_clone.items(): #v0.4
if key.endswith(('running_mean', 'running_var')): #v0.4
del model_dict[key]
style_model.load_state_dict(model_dict, True)
style_model.eval()
#if cuda:
style_loader = StyleLoader(style_folder, style_size)
style_model.cuda()
#else:
# style_loader = StyleLoader(style_folder, style_size, False)
# Define the codec and create VideoWriter object
height = demo_size
width = int(4.0/3*demo_size)
swidth = int(width/4)
sheight = int(height/4)
#if record:
# fourcc = cv2.VideoWriter_fourcc('F','M','P','4')
# out = cv2.VideoWriter('output.mp4', fourcc, 20.0, (2*width, height))
cam = cv2.VideoCapture(0)
cam.set(3, width)
cam.set(4, height)
key = 0
idx = 0
while True:
# read frame
idx += 1
ret_val, img = cam.read()
#if mirror:
# img = cv2.flip(img, 1)
#cimg = img.copy()###############################OR-CAM
img = np.array(img).transpose(2, 0, 1)
# changing style
if idx%20 == 1:
style_v = style_loader.get(int(idx/20))
style_v = Variable(style_v.data)
style_model.setTarget(style_v)
img=torch.from_numpy(img).unsqueeze(0).float()
#if cuda:
img=img.cuda()
img = Variable(img)
img = style_model(img)
#print(type(img))
#if cuda:
#simg = style_v.cpu().data[0].numpy()
img=img.clamp(0, 255).data[0].cpu().detach().numpy()
##x_gpu = cp.array()
##cp.asnumpy(x_gpu)
#img=img.data
#img=img.numpy()
#print(type(img))
#else:
# simg = style_v.data.numpy()
# img = img.clamp(0, 255).data[0].numpy()
img = img.transpose(1, 2, 0).astype('uint8')
#simg = simg.transpose(1, 2, 0).astype('uint8')
# display
#simg = cv2.resize(simg,(swidth, sheight), interpolation = cv2.INTER_CUBIC)
#cimg[0:sheight,0:swidth,:]=simg
#img = np.concatenate((cimg,img),axis=1)
img = cv2.resize(img, None, fx=3.0,fy=3.0, interpolation=cv2.INTER_LINEAR)
kernel = np.ones((5,5),np.float32)/25
img = cv2.filter2D(img,-1,kernel)
cv2.imshow('MSG Demo', img)
#cv2.imwrite('stylized/%i.jpg'%idx,img)
key = cv2.waitKey(1)
#if args.record:
# out.write(img)
if key == 27:
break
cam.release()
#if args.record:
# out.release()
cv2.destroyAllWindows()
def run_demo(cuda, record, vfile):
model = 'models/21styles.params'
ngf = 128
style_size = 512
style_folder = 'images/styles/'
mirror = False
vDir = './video/'
vPath = vDir + vfile
oFile = 'output21-' + vfile
wM, hM = 640, 480
if cuda:
ctx = mx.gpu(0)
os.environ['MXNET_CUDNN_AUTOTUNE_DEFAULT'] = '0'
else:
ctx = mx.cpu(0)
style_loader = StyleLoader(style_folder, style_size, ctx)
style_model = Net(ngf=ngf)
style_model.load_parameters(model, ctx=ctx)
metadata = ffprobe(vPath)
fps = metadata["video"]["@avg_frame_rate"]
# print(json.dumps(metadata["video"], indent=4))
w, h = int(metadata["video"]["@width"]), int(metadata["video"]["@height"])
downsize = h > hM
if downsize:
w = 2 * int(w * hM / h / 2)
h = hM
# downsize = w > wM
# if downsize :
# h = 2 * int(h * wM / w / 2); w = wM
swidth = int(w / 4)
sheight = int(h / 4)
wName = vfile + ' STYLIZED VIDEO fps:' + fps + ' W:' + str(
w) + ' H:' + str(h)
if record:
out = FFmpegWriter(vDir + oFile,
inputdict={
'-r': str(fps),
'-s': '{}x{}'.format(2 * w, h)
},
outputdict={
'-r': str(fps),
'-c:v': 'h264'
})
key, idx = 0, 0
cv2.namedWindow(wName, cv2.WINDOW_NORMAL)
cv2.resizeWindow(wName, 2 * w, h)
for img in vreader(vPath):
idx += 1
if downsize:
img = cv2.resize(img, (w, h), interpolation=cv2.INTER_AREA)
if mirror:
img = cv2.flip(img, 1)
cimg = img.copy()
img = np.array(img).transpose(2, 0, 1).astype(float)
img = F.expand_dims(mx.nd.array(img, ctx=ctx), 0)
# changing styles
if idx % 50 == 1:
style_v = style_loader.get(int(idx / 20))
style_model.set_target(style_v)
img = style_model(img)
simg = np.squeeze(style_v.asnumpy())
simg = simg.transpose(1, 2, 0).astype('uint8')
img = F.clip(img[0], 0, 255).asnumpy()
img = img.transpose(1, 2, 0).astype('uint8')
# display
simg = cv2.resize(simg, (swidth, sheight),
interpolation=cv2.INTER_CUBIC)
cimg[0:sheight, 0:swidth, :] = simg
img = np.concatenate((cimg, cv2.cvtColor(img, cv2.COLOR_BGR2RGB)),
axis=1)
if record:
out.writeFrame(img)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
cv2.imshow(wName, img)
key = cv2.waitKey(1)
if key == 27: # Esc
break
if record:
out.close()
transferAudio(vPath, vDir, oFile)
print("Done OK. Created Stylised Video file", vDir + oFile)
print("fps :", fps, " W:", w, " H:", h)
cv2.destroyAllWindows()
def run_demo(args, mirror=False):
# style_model = Net(ngf=args.ngf)
# style_model.load_state_dict(torch.load(args.model))
model_dict = torch.load(args.model)
model_dict_clone = model_dict.copy() # We can't mutate while iterating
for key, value in model_dict_clone.items():
if key.endswith(('running_mean', 'running_var')):
del model_dict[key]
style_model = Net(ngf=args.ngf)
style_model.load_state_dict(model_dict, False)
style_model.eval()
if args.cuda:
style_loader = StyleLoader(args.style_folder, args.style_size)
style_model.cuda()
else:
style_loader = StyleLoader(args.style_folder, args.style_size, False)
# Define the codec and create VideoWriter object
height = args.demo_size
width = int(4.0 / 3 * args.demo_size)
swidth = int(width / 4)
sheight = int(height / 4)
#lxy
if args.video == '':
videofile = '/home/xyliu/Videos/sports/dance.mp4'
else:
videofile = args.video
print('handle video\'s path is', videofile)
cam = cv2.VideoCapture(videofile)
video_length = int(cam.get(cv2.CAP_PROP_FRAME_COUNT))
ret_val, img = cam.read()
i = 1
## output name
if args.output == '':
dirname = os.path.dirname(videofile)
basename = os.path.basename(videofile)
output_name = os.path.join(dirname, 'tra_' + basename)
else:
output_name = args.output
if args.record:
fshape = img.shape
fheight, fwidth = fshape[:2]
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
# out = cv2.VideoWriter('new_output.mp4', fourcc, 20.0, (2*width, height))
out = cv2.VideoWriter(output_name, fourcc, 20.0, (fwidth * 2, fheight))
cam.set(3, width)
cam.set(4, height)
key = 0
idx = 0
while (cam.isOpened()) and ret_val == True and i < video_length:
# read frame
idx += 1
ret_val, img = cam.read()
i += 1
if mirror:
img = cv2.flip(img, 1)
cimg = img.copy()
img = np.array(img).transpose(2, 0, 1)
# changing style
if idx % 20 == 1:
style_v = style_loader.get(int(idx / 20))
style_v = Variable(style_v.data)
# print('style change to ', style_v)
style_model.setTarget(style_v)
img = torch.from_numpy(img).unsqueeze(0).float()
if args.cuda:
img = img.cuda()
img = Variable(img)
## handle image
img = style_model(img)
if args.cuda:
simg = style_v.cpu().data[0].numpy()
img = img.cpu().clamp(0, 255).data[0].numpy()
else:
simg = style_v.data().numpy()
img = img.clamp(0, 255).data[0].numpy()
img = img.transpose(1, 2, 0).astype('uint8')
simg = simg.transpose(1, 2, 0).astype('uint8')
# display
simg = cv2.resize(simg, (swidth, sheight),
interpolation=cv2.INTER_CUBIC)
cimg[0:sheight, 0:swidth, :] = simg
img = np.concatenate((cimg, img), axis=1)
cv2.imshow('MSG Demo', img)
#cv2.imwrite('stylized/%i.jpg'%idx,img)
key = cv2.waitKey(1)
if args.record:
out.write(img)
if key == 27:
break
cam.release()
if args.record:
out.release()
cv2.destroyAllWindows()
print('video save in ', output_name)
def run_demo(args, mirror=False):
style_model = Net(ngf=args.ngf)
# loading trained model specified by user
style_model.load_state_dict(torch.load(args.model))
style_model.eval()
# processing the style specified and applying it to image
if args.cuda:
style_loader = StyleLoader(args.style_folder, args.style_size)
style_model.cuda()
else:
style_loader = StyleLoader(args.style_folder, args.style_size, False)
# Define the codec and create VideoWriter object
# creating another adjacent video
height = args.demo_size
width = int(4.0/3*args.demo_size)
swidth = int(width/4)
sheight = int(height/4)
# if I put option record, it will save my video
if args.record:
fourcc = cv2.VideoWriter_fourcc('F','M','P','4')
out = cv2.VideoWriter('output.mp4', fourcc, 20.0, (2*width, height))
# capture video camera
cam = cv2.VideoCapture(0)
cam.set(3, width)
cam.set(4, height)
key = 0
idx = 0
# read it frame by frame, and flip image to other side and apply style
while True:
# read frame
idx += 1
ret_val, img = cam.read()
if mirror:
img = cv2.flip(img, 1)
cimg = img.copy()
img = np.array(img).transpose(2, 0, 1)
# changing style
if idx%20 == 1:
style_v = style_loader.get(int(idx/20))
style_v = Variable(style_v.data)
style_model.setTarget(style_v)
img=torch.from_numpy(img).unsqueeze(0).float()
if args.cuda:
img=img.cuda()
img = Variable(img)
img = style_model(img)
if not args.cuda:
simg = style_v.cpu().data[0].numpy()
img = img.cpu().clamp(0, 255).data[0].numpy()
else:
simg = style_v.data().numpy()
img = img.clamp(0, 255).data[0].numpy()
img = img.transpose(1, 2, 0).astype('uint8')
simg = simg.transpose(1, 2, 0).astype('uint8')
# display
simg = cv2.resize(simg,(swidth, sheight), interpolation = cv2.INTER_CUBIC)
cimg[0:sheight,0:swidth,:]=simg
img = np.concatenate((cimg,img),axis=1)
cv2.imshow('MSG Demo', img)
#cv2.imwrite('stylized/%i.jpg'%idx,img)
key = cv2.waitKey(1)
if args.record:
out.write(img)
if key == 27:
break
cam.release()
if args.record:
out.release()
cv2.destroyAllWindows()