def do_processing(path_to_file):
# Choose which CPU or GPU card to use
dev = mx.cpu()
ctx = dev
# Params
long_edge = args.long_edge
# Load data
dir_sketch = "data/sk/"
file_sketch = os.listdir(dir_sketch)
file_sketch.sort()
training = int(0.7 * len(file_sketch)) # 70%
num_sketch = int(len(file_sketch) - training) # 30%
logging.info("Num of sketches: %d" % num_sketch)
print(dir_sketch + file_sketch[training])
# Init sketch
sketch_np = data_processing.PreprocessSketchImage(path_to_file, long_edge)
logging.info("load the sketch image, size = %s", sketch_np.shape[2:])
dshape = sketch_np.shape
clip_norm = 0.05 * np.prod(dshape)
# Load pretrained params
gens = dcnn.get_module("g0", dshape, ctx)
gens.load_params("model/0265-0011975-sketch-vgg16.params")
# Testing
logging.info('Start testing arguments %s', args)
# Load sketch
sketch_data = []
path_sketch = dir_sketch + file_sketch[training]
sketch_np = data_processing.PreprocessSketchImage(path_to_file, long_edge)
mySketch = mx.nd.array(sketch_np)
sketch_data.append(mySketch)
gens.forward(mx.io.DataBatch([sketch_data[-1]], [0]), is_train=False)
new_img = gens.get_outputs()[0]
#cv2.namedWindow('output', cv2.WINDOW_NORMAL)
#new_image=new_img.asnumpy()
#print(new_image)
print(file_sketch[training])
#cv2.destroyAllWindows()
data_processing.SaveImage(new_img.asnumpy(),
"output/test/out_%s" % file_sketch[training])
new_image = cv2.imread("output/test/out_%s" % file_sketch[training])
cv2.imshow('img', new_image)
cv2.waitKey(10)
def do_processing(path_to_file):
# Choose which CPU or GPU card to use
dev = mx.gpu(args.gpu) if args.gpu >= 0 else mx.cpu()
ctx = dev
# Params
long_edge = args.long_edge
# load data
img = cv2.imread(path_to_file)
img_path = "/home/jo/mainproject/sketch/data/sk/images1.jpg"
# Init sketch
sketch_np = data_processing.PreprocessSketchImage(img_path, long_edge)
print("Colored")
# sketch_np = data_processing.PreprocessSketchImage(dir_sketch + file_sketch[training], long_edge)
logging.info("load the sketch image, size = %s", sketch_np.shape[2:])
dshape = sketch_np.shape
clip_norm = 0.05 * np.prod(dshape)
# Load pretrained params
gens = dcnn.get_module("g0", dshape, ctx)
gens.load_params("model/0020-0000088-sketch.params")
#Testing
sketch_data = []
sketch_np = data_processing.PreprocessSketchImage(img, long_edge)
# mySketch = mx.nd.array(sketch_np)
# sketch_data.append(mySketch)
gens.forward(mx.io.DataBatch([sketch_data[-1]], [0]), is_train=False)
new_img = gens.get_outputs()[0]
# output = data_processing.SaveImage(new_img.asnumpy(),"output/test/out_%s" % file_sketch[training+idx])
output = data_processing.SaveImage(new_img.asnumpy(), "output/test/out_%s")
cv2.imwrite("output.jpg", output)
import gen_v4
dshape = (1, 3, 480, 640)
clip_norm = 1.0 * np.prod(dshape)
model_prefix = "./model/"
ctx = mx.gpu(0)
# generator
gens = [
gen_v4.get_module("g0", dshape, ctx),
gen_v3.get_module("g1", dshape, ctx),
gen_v3.get_module("g2", dshape, ctx),
gen_v4.get_module("g3", dshape, ctx)
]
for i in range(len(gens)):
gens[i].load_params("./model/%d/v3_0002-0026000.params" % i)
content_np = data_processing.PreprocessContentImage("../IMG_4343.jpg",
min(dshape[2:]), dshape)
data = [mx.nd.array(content_np)]
for i in range(len(gens)):
gens[i].forward(mx.io.DataBatch([data[-1]], [0]), is_train=False)
new_img = gens[i].get_outputs()[0]
data.append(new_img.copyto(mx.cpu()))
data_processing.SaveImage(new_img.asnumpy(), "out_%d.jpg" % i)
import os
os.system("rm -rf out.zip")
os.system("zip out.zip out_*")
grad = mx.nd.zeros(sketch_nd.shape)
tv_grad_executor = get_tv_grad_executor(gens.get_outputs()[0],
ctx, args.tv_weight)
tv_grad_executor.forward()
# new_grad
grad[:] += loss_grad_array + tv_grad_executor.outputs[0].copyto(mx.cpu())
gnorm = mx.nd.norm(grad).asscalar()
if gnorm > clip_norm:
#logging.info("Epoch[%d] Image[%d]: Data Grad: %.5f" %\
# (i, idx, (gnorm / clip_norm)))
grad[:] *= clip_norm / gnorm
tot_gnorm += gnorm / clip_norm
gens.backward([grad])
gens.update()
new_img = gens.get_outputs()[0]
if (idx+1) % num_sketch == 0:
ave_gnorm = tot_gnorm / num_sketch
logging.info("Epoch[%d]: Epoch Data Grad: %.5f" % (i+1, ave_gnorm))
if ((i+1) % args.save_epochs == 0 and (idx+1) % num_sketch == 0):
data_processing.SaveImage(new_img.asnumpy(), 'output/train/out_'+str(i+1)+'_'+file_sketch[idx])
gens.save_params("model/%04d-%07d-sketch.params" % ((i+1), (idx+1)))
logging.info("Save Params : %04d-%07d-sketch.params" % ((i+1), (idx+1)))
logging.info("Time elapsed for one epoch in seconds: " + str(time.time()-tic))
file_sketch.sort()
training = int(0.7 * len(file_sketch)) # 70%
num_sketch = int(len(file_sketch) - training) # 30%
logging.info("Num of sketches: %d" % num_sketch)
# Init sketch
sketch_np = data_processing.PreprocessSketchImage(
dir_sketch + file_sketch[training], long_edge)
logging.info("load the sketch image, size = %s", sketch_np.shape[2:])
dshape = sketch_np.shape
clip_norm = 0.05 * np.prod(dshape)
# Load pretrained params
gens = dcnn.get_module("g0", dshape, ctx)
gens.load_params("model/0265-0011975-sketch-vgg16.params")
# Testing
logging.info('Start testing arguments %s', args)
for idx in range(num_sketch):
# Load sketch
sketch_data = []
path_sketch = dir_sketch + file_sketch[training + idx]
sketch_np = data_processing.PreprocessSketchImage(path_sketch, long_edge)
mySketch = mx.nd.array(sketch_np)
sketch_data.append(mySketch)
gens.forward(mx.io.DataBatch([sketch_data[-1]], [0]), is_train=False)
new_img = gens.get_outputs()[0]
data_processing.SaveImage(
new_img.asnumpy(), "output/test/out_%s" % file_sketch[training + idx])