1. img = cv2.imread(image_path)   # cv2.imread() 读出来的是BGR图片，而一般的读图片RGB，如imageFolder，dataloader

2.  

3. cv2.imwrite('image/gakki.png',img)    # image 前面不能加 /   不然写不出来

4. img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)   # 转为 RGB

或者 

方法一

b,g,r=cv2.split(img)

img2=cv2.merge([r.g.b])

方法二

img2=img[: , : , : : -1] = img[...,::-1]  = img[...,[2,0,1]]

另外： 

im_color = cv2.applyColorMap(im_gray or img, cv2.COLORMAP_JET)

COLORMAP_AUTUMN = 0,
COLORMAP_BONE = 1,
COLORMAP_JET = 2,
COLORMAP_WINTER = 3,
COLORMAP_RAINBOW = 4,
COLORMAP_OCEAN = 5,
COLORMAP_SUMMER = 6,
COLORMAP_SPRING = 7,
COLORMAP_COOL = 8,
COLORMAP_HSV = 9,
COLORMAP_PINK = 10,
COLORMAP_HOT = 11
https://blog.****.net/JNingWei/article/details/78803669

另外在做CAM时候，

cam = np.dot(weight[pred_class],(feature_conv[i].reshape((num_channel, h * w))))
cam = cam.reshape((h, w))
cam = cam - np.min(cam)
cam_img = cam / np.max(cam)
cam_img = np.uint8(255 * cam_img)
cam_img = cv2.resize(cam_img, (448, 448))
cam_img = cv2.applyColorMap(cam_img, cv2.COLORMAP_JET)

由于从14*14生成的，cv2本身得到的 448*448*3 就为 bgr 模式， 直接 cv2.imwrite（）就可以了

np.expand_dims(a,0).repeat(1000,axis=0)   # numpy扩展维度