【发布时间】:2019-01-01 17:34:27
【问题描述】:
我正在使用 opencv 进行一些神经风格传输技巧,但我无法将视频保存到文件中。该文件已创建,但只有 6 kb 大。
from imutils.video import VideoStream
from imutils import paths
import itertools
import argparse
import imutils
import time
import cv2
import numpy as np
ap = argparse.ArgumentParser()
ap.add_argument("-m", "--models", required=True,
help="path to directory containing neural style transfer models")
args = vars(ap.parse_args())
modelPaths = paths.list_files(args["models"], validExts=(".t7",))
modelPaths = sorted(list(modelPaths))
models = list(zip(range(0, len(modelPaths)), (modelPaths)))
modelIter = itertools.cycle(models)
(modelID, modelPath) = next(modelIter)
print("[INFO] loading style transfer model...")
net = cv2.dnn.readNetFromTorch(modelPath)
print("[INFO] starting video stream...")
vs = VideoStream(src=0).start()
time.sleep(2.0)
print("[INFO] {}. {}".format(modelID + 1, modelPath))
out = cv2.VideoWriter('output2.avi', cv2.VideoWriter_fourcc(*'XVID'), 5, (451, 600))
while True:
frame = vs.read()
frame = imutils.resize(frame, width=600)
orig = frame.copy()
(h, w) = frame.shape[:2]
blob = cv2.dnn.blobFromImage(frame, 1.0, (w, h),
(103.939, 116.779, 123.680), swapRB=False, crop=False)
net.setInput(blob)
output = net.forward()
output = output.reshape((3, output.shape[2], output.shape[3]))
output[0] += 103.939
output[1] += 116.779
output[2] += 123.680
output /= 255.0
output = output.transpose(1, 2, 0)
out.write(np.uint8(output))
cv2.imshow("Input", frame)
cv2.imshow("Output", output)
key = cv2.waitKey(1) & 0xFF
if key == ord("n"):
(modelID, modelPath) = next(modelIter)
print("[INFO] {}. {}".format(modelID + 1, modelPath))
net = cv2.dnn.readNetFromTorch(modelPath)
print(frame.shape)
elif key == ord("q"):
break
cv2.destroyAllWindows()
vs.stop()
主要的事情发生在:
out = cv2.VideoWriter('output2.avi', cv2.VideoWriter_fourcc(*'XVID'), 5, (450, 600))
output = net.forward() # which computes the neural styled output image
为了将输出/帧保存到文件,我做了:
out.write(np.uint8(output))
关于为什么我的代码不起作用的任何想法?我尝试了许多编解码器和文件类型组合,但我认为问题不存在。大家觉得可能是次元吗? (我现在有 450、600,因为我打印了 output.shape,它返回了 450,600,3,所以我认为 450 x 600 听起来不错)。
【问题讨论】:
-
如果你在视频中写
np.uint8(output),那么你应该imshow同样的事情。我看到output /= 255.0,所以我猜输出是一个浮点数组,其值在 [0.0, 1.0] 范围内。将其转换为uint8会将其减少到仅 0 和 1。但是,imshow(根据文档)会将浮点图像转换并缩放回 0-255 范围。 -
@DanMašek 是正确的。你可以阅读更多关于他描述的内容here。
-
确保您的应用程序可以访问 opencv_ffmpeg dll(或共享库?)
标签: python opencv conv-neural-network codec