I turned the np functions to gpu functions that works with pycuda.

My last problem is the binarize problem. I have to change the opencv functions cpu to the gpu.

But python opencv functions work for only cpu. So this is a real problem. I have a 500 line code and this is my last problem on this code, lol. What do you should for me?

Here is the code:

```
frame = np.copy(frame)## ayni degiskene ayni dosyayi farkli formatta okutturdum
hls = cv2.cvtColor(frame, cv2.COLOR_RGB2HLS).astype(np.float) ##HLS renk formatina cevirdik ve kanallari ayistirdim.
l_channel = hls[:,:,1]
s_channel = hls[:,:,2]
sobelx = cv2.Sobel(l_channel, cv2.CV_64F, 1, 0) ##Gorselin L kanalindaki kisminin turevini x olarak aldim.
abs_sobelx = np.absolute(sobelx) ##Satirlari dikey olarak hesaplayacagiz bu yuzden yatay olarak almasini engellemek icin turevin sonucunu mutlak deger aldik.
scaled_sobel = np.uint8(255*abs_sobelx/np.max(abs_sobelx))
##Egimini hesapladim.
sxbinary = np.zeros_like(scaled_sobel)
sxbinary[(scaled_sobel >= sx_thresh[0]) & (scaled_sobel <= sx_thresh[1])] = 1
##Goruntunun doygunluk durumunu ogrenmek icin kanallari ayristirarak Threshold uyguladim.
s_binary = np.zeros_like(s_channel)
s_binary[(s_channel >= s_thresh[0]) & (s_channel <= s_thresh[1])] = 1
##Isik durumunu ogrenmek icin bi ustteki islemi uyguladim
l_binary = np.zeros_like(l_channel)
l_binary[(l_channel >= l_thresh[0]) & (l_channel <= l_thresh[1])] = 1
channels = 255*np.dstack(( l_binary, sxbinary, s_binary)).astype('uint8')
binary = np.zeros_like(sxbinary)
binary[((l_binary == 1) & (s_binary == 1) | (sxbinary==1))] = 1
binary = 255*np.dstack((binary,binary,binary)).astype('uint8')
return binary,channels
```