def main():
# location of depth module, config and parameters
model_name = 'depthnormals_nyud_alexnet'
#model_name = 'depthnormals_nyud_vgg'
# module_fn = 'models/iccv15/%s.py' % model_name
# config_fn = 'models/iccv15/%s.conf' % model_name
# params_dir = 'weights/iccv15/%s' % model_name
module_fn = 'models/iccv15/depthnormals_nyud_alexnet.py' #% model_name
config_fn = 'models/iccv15/depthnormals_nyud_alexnet.conf' #% model_name
params_dir = 'weights/iccv15/depthnormals_nyud_alexnet' #% model_name
# load depth network
machine = net.create_machine(module_fn, config_fn, params_dir)
# demo image
rgb = Image.open('thyroid.jpg')
rgb = rgb.resize((320, 240), Image.BICUBIC)
#rergb = Image.fromarray(rgb)
rgb.save("thyroid.jpg")
# build depth inference function and run
rgb_imgs = np.asarray(rgb).reshape((1, 240, 320, 3))
(pred_depths, pred_normals) = machine.infer_depth_and_normals(rgb_imgs)
# save prediction
depth_img_np = depth_montage(pred_depths)
depth_img = Image.fromarray((255 * depth_img_np).astype(np.uint8))
depth_img.save('demo_depth_prediction_thyroid.png')
normals_img_np = normals_montage(pred_normals)
normals_img = Image.fromarray((255 * normals_img_np).astype(np.uint8))
normals_img.save('demo_normals_prediction_thyroid.png')
def main():
# location of depth module, config and parameters
model_name = 'depthnormals_nyud_alexnet'
#model_name = 'depthnormals_nyud_vgg'
# module_fn = 'models/iccv15/%s.py' % model_name
# config_fn = 'models/iccv15/%s.conf' % model_name
# params_dir = 'weights/iccv15/%s' % model_name
module_fn = 'models/iccv15/depthnormals_nyud_alexnet.py' #% model_name
config_fn = 'models/iccv15/depthnormals_nyud_alexnet.conf' #% model_name
params_dir = 'weights/iccv15/depthnormals_nyud_alexnet' #% model_name
# load depth network
machine = net.create_machine(module_fn, config_fn, params_dir)
# demo image
rgb = Image.open('thyroid.jpg')
rgb = rgb.resize((320, 240), Image.BICUBIC)
#rergb = Image.fromarray(rgb)
rgb.save("thyroid.jpg")
# build depth inference function and run
rgb_imgs = np.asarray(rgb).reshape((1, 240, 320, 3))
(pred_depths, pred_normals) = machine.infer_depth_and_normals(rgb_imgs)
# save prediction
depth_img_np = depth_montage(pred_depths)
depth_img = Image.fromarray((255*depth_img_np).astype(np.uint8))
depth_img.save('demo_depth_prediction_thyroid.png')
normals_img_np = normals_montage(pred_normals)
normals_img = Image.fromarray((255*normals_img_np).astype(np.uint8))
normals_img.save('demo_normals_prediction_thyroid.png')
def main():
# location of depth module, config and parameters
module_fn = 'models/depth_kitti.py'
config_fn = 'models/depth_kitti.conf'
params_dir = 'weights/depth_kitti'
# load depth network
machine = net.create_machine(module_fn, config_fn, params_dir)
# demo image
rgb = Image.open('demo_kitti_rgb.jpg')
rgb = rgb.resize((612, 184), Image.BICUBIC)
# build depth inference function and run
rgb_imgs = np.asarray(rgb).reshape((1, 184, 612, 3))
pred_depths = machine.infer_depth(rgb_imgs)
# save prediction
(m, M) = (pred_depths.min(), pred_depths.max())
depth_img_np = (pred_depths[0] - m) / (M - m)
depth_img = Image.fromarray((255 * depth_img_np).astype(np.uint8))
depth_img.save('demo_kitti_depth_prediction.png')
import matplotlib.pyplot as plt
plt.ion()
import ipdb
ipdb.set_trace()
pass
def machine2model():
module_fn = 'models/iccv15/depthnormals_nyud_alexnet.py' #% model_name
config_fn = 'models/iccv15/depthnormals_nyud_alexnet.conf' #% model_name
params_dir = 'weights/iccv15/depthnormals_nyud_alexnet' #% model_name
trainMachine = net.create_machine(module_fn, config_fn, params_dir)
#model = trainMachine.define_machine()
print trainMachine
return trainMachine
def machine2model():
module_fn = 'models/iccv15/depthnormals_nyud_alexnet.py' #% model_name
config_fn = 'models/iccv15/depthnormals_nyud_alexnet.conf' #% model_name
params_dir = 'weights/iccv15/depthnormals_nyud_alexnet' #% model_name
trainMachine = net.create_machine(module_fn, config_fn, params_dir)
#model = trainMachine.define_machine()
print trainMachine
return trainMachine
def load_depth_nn(
model_name="depthnormals_nyud_vgg", nn_path="/afs/csail.mit.edu/u/y/yuvval/externals/dnl-depthnormals/"
):
# model_name = 'depthnormals_nyud_alexnet' # an alternative model_name
# add path of the neural network
sys.path.append(nn_path)
from utils import depth_montage, normals_montage
import net
depth_nn = dict()
# location of depth module, config and parameters
depth_nn["model_name"] = model_name
depth_nn["module_fn"] = "%smodels/iccv15/%s.py" % (nn_path, model_name)
depth_nn["config_fn"] = "%smodels/iccv15/%s.conf" % (nn_path, model_name)
depth_nn["params_dir"] = "%sweights/iccv15/%s" % (nn_path, model_name)
# load depth network
depth_nn["machine"] = net.create_machine(depth_nn["module_fn"], depth_nn["config_fn"], depth_nn["params_dir"])
return depth_nn
def load_depth_nn(
model_name='depthnormals_nyud_vgg',
nn_path='/afs/csail.mit.edu/u/y/yuvval/externals/dnl-depthnormals/'):
#model_name = 'depthnormals_nyud_alexnet' # an alternative model_name
# add path of the neural network
sys.path.append(nn_path)
from utils import depth_montage, normals_montage
import net
depth_nn = dict()
# location of depth module, config and parameters
depth_nn['model_name'] = model_name
depth_nn['module_fn'] = '%smodels/iccv15/%s.py' % (nn_path, model_name)
depth_nn['config_fn'] = '%smodels/iccv15/%s.conf' % (nn_path, model_name)
depth_nn['params_dir'] = '%sweights/iccv15/%s' % (nn_path, model_name)
# load depth network
depth_nn['machine'] = net.create_machine(depth_nn['module_fn'],
depth_nn['config_fn'],
depth_nn['params_dir'])
return depth_nn
def main():
# location of depth module, config and parameters
module_fn = 'models/depth.py'
config_fn = 'models/depth.conf' #网络结构
params_dir = 'weights/depth' #网络相关参数
# load depth network
machine = net.create_machine(module_fn, config_fn, params_dir)
# demo image
rgb = Image.open('demo_nyud_rgb.jpg')
rgb = rgb.resize((320, 240), Image.BICUBIC)
# build depth inference function and run
rgb_imgs = np.asarray(rgb).reshape((1, 240, 320, 3))
pred_depths = machine.infer_depth(rgb_imgs)
# save prediction
(m, M) = (pred_depths.min(), pred_depths.max())
depth_img_np = (pred_depths[0] - m) / (M - m)
depth_img = Image.fromarray((255 * depth_img_np).astype(np.uint8))
depth_img.save('demo_nyud_depth_prediction.png')
def main():
# location of depth module, config and parameters
module_fn = 'models/depth.py'
config_fn = 'models/depth.conf'#网络结构
params_dir = 'weights/depth'#网络相关参数
# load depth network
machine = net.create_machine(module_fn, config_fn, params_dir)
# demo image
rgb = Image.open('demo_nyud_rgb.jpg')
rgb = rgb.resize((320, 240), Image.BICUBIC)
# build depth inference function and run
rgb_imgs = np.asarray(rgb).reshape((1, 240, 320, 3))
pred_depths = machine.infer_depth(rgb_imgs)
# save prediction
(m, M) = (pred_depths.min(), pred_depths.max())
depth_img_np = (pred_depths[0] - m) / (M - m)
depth_img = Image.fromarray((255*depth_img_np).astype(np.uint8))
depth_img.save('demo_nyud_depth_prediction.png')
import os
import sys
import numpy as np
from PIL import Image
import net
def predictDepth(filelist, pwd):
image_w = 416
image_h = 218
# location of depth module, config and parameters
module_fn = pwd + '/eigen/models/depth_kitti.py'
config_fn = pwd + '/eigen/models/depth_kitti.conf'
params_dir = pwd + '/eigen/weights/depth_kitti'
# load depth network
machine = net.create_machine(module_fn, config_fn, params_dir)
rgb_imgs = []
# demo image
for filename in filelist:
rgb = Image.open(filename)
rgb = rgb.resize((image_w, image_h), Image.BICUBIC)
rgb_imgs.append(rgb)
# build depth inference function and run
rgb_imgs = np.asarray(rgb_imgs).reshape((len(filelist), image_h, image_w, 3))
pred_depths = machine.infer_depth(rgb_imgs)
# save prediction
# (m, M) = (pred_depths.min(), pred_depths.max())
# depth_img_np = (pred_depths[0] - m) / (M - m)
np.save(pwd + "/eigen/output/depth.npy", pred_depths)
return pred_depths
