牛津视觉图书馆
我在录制Kinect流的简单视频之上使用EmguCV。
I'm using EmguCV on top of simple video recorded of Kinect stream.
是否有最佳实践在视频上使用 牛津视觉图书馆而无需使用服务器吗？
Is there best practicies to use Oxford Vision Library on Video without spaming the server ?
- 在查询服务器之前可能是一个很好的离线算法？
- 或上传视频在Azure然后运行牛津视觉代码？ （安全性，带宽等......）
- may be good offline algorithm, before querying the server ?
- or uploading the video on Azure then running a Oxford vision code ? (what about security, bandwidth, ...)
你好JP，
一个您今天可以采取的方法是定期捕获静态图像并将其发送到Computer Vision API。
One approach you can take today is to capture still images periodically and send those to the Computer Vision API.
计算机视觉API受限于每分钟20次通话，因此您可以每隔3秒向API提交一张静止图像。我见过的另一种方法是发送图像的频率低于每次发送图像（例如每10秒发送一次），然后在检测到某个
状态变化后更频繁地捕获图像。
The Computer Vision API is throttled at 20 calls per minute, so you can submit a still image every 3 seconds to the API. Another approach that I've seen used is to send images less frequently than that (say one every 10 seconds) and then after a certain state change is detected to capture images more frequently.
另外一个提示是，您可以在通过线路发送之前缩小图像的大小，因为这可以提高性能。
One additional tip is that you can scale down the size of the image before sending over the wire, as that may improve performance.
感谢您提供有关设置的信息，我们会将此信息反馈给团队，因为我们会优先处理下一组更新。
Thank you for the info on your setup, I'll take this back to the team as we prioritize our next set of updates.