dim_z = input_z.shape.as_list()[1]
vocab_size = input_y.shape.as_list()[1]
initializer = tf.global_variables_initializer()
sess = tf.Session()
sess.run(initializer)
return sess, vocab_size, dim_z, input_z, input_y, input_trunc, output
@biggan.command('sample', inputs={'truncation': 'float', 'category': 'text', 'seed': 'integer'}, outputs={'generatedOutput': 'image', 'z': 'vector'})
def sample_cmd(model, inp):
(sess, vocab_size, dim_z, input_z, input_y, input_trunc, output) = model
z = truncated_z_sample(1, dim_z, inp['truncation'], seed=inp['seed'])
y = CATEGORIES.index(inp['category'])
ims = sample(sess, z, y, vocab_size, input_z, input_y, input_trunc, output)
return dict(generatedOutput=ims[0], z=z[0])
@biggan.command('generateFromVector', inputs={'z': 'vector', 'category': 'text'}, outputs={'generatedOutput': 'image'})
def generate_from_vector(model, inp):
(sess, vocab_size, dim_z, input_z, input_y, input_trunc, output) = model
z = inp['z'].reshape(-1, inp['z'].shape[0])
y = CATEGORIES.index(inp['category'])
ims = sample(sess, z, y, vocab_size, input_z, input_y, input_trunc, output)
return dict(generatedOutput=ims[0])
if __name__ == '__main__':
biggan.run()
model = 'checkpoints/karras2019stylegan-ffhq-1024x1024.pkl'
print("open model %s" % model)
with open(model, 'rb') as file:
G, D, Gs = pickle.load(file)
return Gs
@stylegan.command('convert',
inputs={
'z': 'vector',
'truncation': 'float'
},
outputs={'output': 'image'})
def convert(Gs, inp):
truncation = inp['truncation']
latents = np.array(inp['z']).reshape(
(1,
512)) # np.random.RandomState(1000).randn(1, *Gs.input_shapes[0][1:])
#labels = np.zeros([latents.shape[0]] + Gs.input_shapes[1][1:])
images = Gs.run(latents,
None,
truncation_psi=truncation,
randomize_noise=False,
output_transform=fmt)
output = np.clip(images[0], 0, 255).astype(np.uint8)
return dict(output=output)
if __name__ == '__main__':
stylegan.run()
from maskrcnn_benchmark.config import cfg
from predictor import COCODemo
import numpy as np
import torch
from runway import RunwayModel
masrcnn = RunwayModel()
@masrcnn.setup
def setup():
config_file = "configs/caffe2/e2e_mask_rcnn_R_50_FPN_1x_caffe2.yaml"
cfg.merge_from_file(config_file)
cfg.merge_from_list(["MODEL.DEVICE", "cpu"])
coco_demo = COCODemo(
cfg,
min_image_size=800,
confidence_threshold=0.7,
)
return coco_demo
@masrcnn.command('mask', inputs={'image': 'image'}, outputs={'image': 'image'})
def mask(model, inp):
img = np.array(inp['image'])
output = coco_demo.run_on_opencv_image(img)
return dict(image=output)
if __name__ == "__main__":
masrcnn.run()
h, w = 480, 640
img_shape = (h, w, 3)
batch_shape = (1, ) + img_shape
g = tf.Graph()
g.as_default()
g.device('/gpu:0')
soft_config = tf.ConfigProto(allow_soft_placement=True)
soft_config.gpu_options.allow_growth = True
sess = tf.Session(config=soft_config)
img_placeholder = tf.placeholder(tf.float32,
shape=batch_shape,
name='img_placeholder')
preds = transform.net(img_placeholder)
load_checkpoint(models[idx_model]["ckpt"], sess)
return sess
@faststyletransfer.command('convert',
inputs={'image': 'image'},
outputs={'output': 'image'})
def upscale(sess, inp):
img = np.array(inp['image'])
img = np.expand_dims(img, 0)
output = sess.run(preds, feed_dict={img_placeholder: img})
output = np.clip(output[0], 0, 255).astype(np.uint8)
return dict(output=output)
if __name__ == '__main__':
faststyletransfer.run()
@pgan.setup
def setup(alpha=0.5):
global Gs
tf.InteractiveSession()
batch_size = 8
model = 'network-snapshot-itpgan.pkl'
print("open model %s" % model)
with open(model, 'rb') as file:
G, D, Gs = pickle.load(file)
return Gs
@pgan.command('convert', inputs={'z': 'vector'}, outputs={'output': 'image'})
def convert(Gs, inp):
latents = np.array(inp['z']).reshape(
(1,
512)) # np.random.RandomState(1000).randn(1, *Gs.input_shapes[0][1:])
labels = np.zeros([latents.shape[0]] + Gs.input_shapes[1][1:])
images = Gs.run(latents, labels)
images = np.clip(np.rint((images + 1.0) / 2.0 * 255.0), 0.0,
255.0).astype(np.uint8) # [-1,1] => [0,255]
images = images.transpose(0, 2, 3, 1) # NCHW => NHWC
output = np.clip(images[0], 0, 255).astype(np.uint8)
return dict(output=output)
if __name__ == '__main__':
pgan.run()
model = infer_engine.initialize_model_from_cfg(weights)
dummy_coco_dataset = dummy_datasets.get_coco_dataset()
return model
@detectron.command('detect',
inputs={'image': 'image'},
outputs={'output': 'image'})
def detect(model, inp):
im = np.array(inp['image'])
with c2_utils.NamedCudaScope(0):
cls_boxes, cls_segms, cls_keyps = infer_engine.im_detect_all(
model, im, None) #, timers=timers)
#results = get_result_json(cls_boxes, cls_segms, cls_keyps, thresh=args.save_thresh, dataset=dummy_coco_dataset)
im_new = vis_utils.vis_one_image_opencv(im[:, :, ::-1],
cls_boxes,
segms=cls_segms,
keypoints=cls_keyps,
thresh=0.9,
kp_thresh=2,
show_box=True,
dataset=dummy_coco_dataset,
show_class=True)
out = np.array(im_new)
output = np.clip(out, 0, 255).astype(np.uint8)
return dict(output=output)
if __name__ == '__main__':
detectron.run()
saver = tf.train.Saver()
path = opts['styleCheckpoint']
model_name = [
p for p in os.listdir(path) if os.path.isdir(os.path.join(path, p))
][0]
checkpoint_dir = os.path.join(path, model_name, 'checkpoint_long')
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
ckpt_name = os.path.basename(ckpt.model_checkpoint_path)
saver.restore(sess, os.path.join(checkpoint_dir, ckpt_name))
return dict(sess=sess, input_photo=input_photo, output_photo=output_photo)
@st.command('stylize',
inputs={'contentImage': 'image'},
outputs={'stylizedImage': 'image'})
def stylize(model, inp):
img = inp['contentImage']
img = np.array(img)
img = img / 127.5 - 1.
img = np.expand_dims(img, axis=0)
img = model['sess'].run(model['output_photo'],
feed_dict={model['input_photo']: img})
img = (img + 1.) * 127.5
img = img.astype('uint8')
img = img[0]
return dict(stylizedImage=img)
if __name__ == '__main__':
st.run()
# launch Runway
glow = RunwayModel()
model_ready = False
@glow.setup
def setup():
global model_ready
print('setup model')
model_ready = True
return None
@glow.command('convert', inputs={'image': 'image'}, outputs={'output': 'image'})
def detect(sess, inp):
img = np.array(inp['image'])
z_addition = 0.95 * z_manipulate[tags.index('Attractive')]
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 2)
if len(rects) == 0 or not model_ready:
print('nothing found')
return dict(output=img)
img = warper.align(img[:, :, ::-1], gray, rects[0], z_addition)[:, :, ::-1]
img = np.array(Image.fromarray(img).convert('RGB'))
output = np.clip(img, 0, 255).astype(np.uint8)
return dict(output=output)
if __name__ == '__main__':
glow.run()
opt.norm = 'batch'
opt.input_nc = 3
opt.output_nc = 1
opt.which_model_netG = 'resnet_9blocks'
opt.no_dropout = True
model = create_model(opt)
return model
@photosketch.command('convert',
inputs={'image': 'image'},
outputs={'output': 'image'})
def convert(model, inp):
img = np.array(inp['image'])
img = img / 255.
h, w = img.shape[0:2]
img = np.transpose(img, (2, 0, 1))
img = np.expand_dims(img, 0)
img = torch.from_numpy(img).float() #.to(device)
data = {'A_paths': '', 'A': img, 'B': img}
model.set_input(data)
model.test()
output = util.tensor2im(model.fake_B)
output = Image.fromarray(output.astype('uint8'))
output = output.convert('RGB')
return dict(output=output)
if __name__ == '__main__':
photosketch.run()