def get_depth(img_path):
parameters = os.path.basename(parameter)
parameters = parameters[0:-10:]
opt = TrainOptions().parse(
) # set CUDA_VISIBLE_DEVICES before import torch
model = create_model(opt, parameters)
pred_depth = test_simple(model, img_path)
return pred_depth
def get_depthmap_img(img_path, save_path, size, checkpoints_dir):
opt = TrainOptions()
#print(opt)
opt.gpu_ids = '0,1'
opt.isTrain = True
opt.checkpoints_dir = checkpoints_dir
opt.name = 'test_local/'
model = create_model(opt)
path = test_simple(model, img_path, size,save_path)
print("We are done")
return(path)
def run(image_in, parameter):
global img_path, parameters
image_in_number = io.imread(image_in) # 从指定位置读取图片
image_in_path = 'MegaDepth/demo_img/demo.jpg'
io.imsave(image_in_path, image_in_number) # 将图片保存到指定文件夹
parameters = os.path.basename(parameter) # 获得参数文件名
parameters = parameters[0:-10:] # 将参数文件名转化为指定文件名
opt = TrainOptions().parse(
) # set CUDA_VISIBLE_DEVICES before import torch
model = create_model(opt, parameters)
pred_depth = test_simple(model) # 获得深度信息
position = 'MegaDepth/demo_img/demo1.jpg'
io.imsave(position, pred_depth) # 保存预测深度图
print("We are done")
return pred_depth, position # 返回深度信息和深度图位置
def get_depthmap_from_list(img_path_list, save_path, checkpoints_dir):
"""
get the depthmap from a list of images paths
"""
opt = TrainOptions()
#print(opt)
opt.gpu_ids = '0,1'
opt.isTrain = True
opt.checkpoints_dir = checkpoints_dir
opt.name = 'test_local/'
paths = []
model = create_model(opt)
for img_path in img_path_list:
size = Image.open(img_path).size
print(size)
path = test_simple(model, img_path, size,save_path)
paths.append(path)
print("We are done")
return(paths)
# visualize prediction using inverse depth, so that we don't need sky segmentation (if you want to use RGB map for visualization, \
# you have to run semantic segmentation to mask the sky first since the depth of sky is random from CNN)
pred_inv_depth = 1/pred_depth
pred_inv_depth = pred_inv_depth.data.cpu().numpy()
# you might also use percentile for better visualization
pred_inv_depth = pred_inv_depth/np.amax(pred_inv_depth)
pred_inv_depth[indices_flip] = np.flip(pred_inv_depth[1:3], 2)
for i, im in enumerate(pred_inv_depth):
pred_inv_depth[i] -= np.min(im)
pred_inv_depth[i] /= np.max(pred_inv_depth[i])
mean_pred = np.mean(pred_inv_depth, axis = 0)
save_filename = os.path.splitext(os.path.basename(img_file))[0] + 'meanpred_depth'
io.imsave(os.path.join(save_path, save_filename + '.jpg'), mean_pred)
np.save(os.path.join(save_path, save_filename + '.npy'), mean_pred )
print("Image saved to "+ os.path.join(save_path, save_filename))
if __name__ == "__main__":
img_path = '../DataSV/milacropresize.jpg'
save_path = '.'
#size = [512,512]
opt = TrainOptions().parse()
model = create_model(opt)
test_simple(model, img_path, save_path)#size,save_path)
print("We are done")
def run():
opt = TrainOptions().parse(
) # set CUDA_VISIBLE_DEVICES before import torch
model = create_model(opt)
test_simple(model)