华为计算微信公众号
昇腾AI开发者公众号
华为计算微博
华为计算今日头条
【问题】
用opencv读视频然后用MindX SDK接口推理，发现opencv库图像预处理比较耗时，打算根据教程用MindX SDK提供的媒体数据处理Image类和ImageProcessor类进行图像预处理，怎么把opencv读取的图片，视频或摄像头图像数据转成Image类。
【需求】尝试代码如下：
import cv2 import numpy as np from mindx.sdk import Tensor, base from mindx.sdk.base import ImageProcessor, Size, Image base.mx_init() # 执行全局初始化后即可正常调用mxVision接口 image_path = "video/world_cup.jpg" image = cv2.imread(image_path) # 读入图片 #img = np.random.randint(0, 255, (640, 400, 3), dtype=np.uint8) image = np.ascontiguousarray(image) print(image.shape, image.dtype) size_para = Size(650, 433) size_para2 = Size(650, 433) iimage = Image(image, base.bgr, (size_para, size_para2)) #iimage(image, base.bgr) iimage.to_host() # 图像解码 # 初始化ImageProcessor对象 # imageProcessor = ImageProcessor(deviceId=0) # 读取图片路径进行解码，解码格式为nv12（YUV_SP_420） # decoded_image = imageProcessor.decode(image_path, base.nv12) image_pre = iimage print(image_pre) # image = cv2.resize(image, (640, 400)) cv2.imshow("Output", image) cv2.waitKey(0) cv2.destroyAllWindows()
我要发帖子
点赞
收藏
回复
分享
举报
已解决 已解决
【问题】
用opencv读视频然后用MindX SDK接口推理，发现opencv库图像预处理比较耗时，打算根据教程用MindX SDK提供的媒体数据处理Image类和ImageProcessor类进行图像预处理，怎么把opencv读取的图片，视频或摄像头图像数据转成Image类。
【需求】尝试代码如下：
import cv2 import numpy as np from mindx.sdk import Tensor, base from mindx.sdk.base import ImageProcessor, Size, Image base.mx_init() # 执行全局初始化后即可正常调用mxVision接口 image_path = "video/world_cup.jpg" image = cv2.imread(image_path) # 读入图片 #img = np.random.randint(0, 255, (640, 400, 3), dtype=np.uint8) image = np.ascontiguousarray(image) print(image.shape, image.dtype) size_para = Size(650, 433) size_para2 = Size(650, 433) iimage = Image(image, base.bgr, (size_para, size_para2)) #iimage(image, base.bgr) iimage.to_host() # 图像解码 # 初始化ImageProcessor对象 # imageProcessor = ImageProcessor(deviceId=0) # 读取图片路径进行解码，解码格式为nv12（YUV_SP_420） # decoded_image = imageProcessor.decode(image_path, base.nv12) image_pre = iimage print(image_pre) # image = cv2.resize(image, (640, 400)) cv2.imshow("Output", image) cv2.waitKey(0) cv2.destroyAllWindows()