Currently we develop the Vision detection program in Suzhou, but the speed is quite slowly. After checked, all data were allocated in US and EU AWS cloud, no option in AP region, like in HK, SG or Japan. Even we are bypass the network, but it still can not get improvement.
Hey @david.kang -
Can you give a little more detail about what parts of the vision offering you are using? Unless you are doing something like passing images to an api that is doing anomaly detection or something, all of the processing for the vision pipeline will be running locally on the device, not on the cloud.
Tulip | Cameras, please find the details in here. We are try to use a Intel RealSense D415 to detect the change. We are follow up the instruction with below link: Getting Started with Vision | Tulip Knowledge Base - Support for Building Operations Apps. But it was challenge to get better performance with network communication.
Hey @david.kang -
Network speed shouldn’t impact vision performance at all because all the heavy lifting is happening on the computer running Tulip player, not some cloud hardware.
What are the specs of the computers you are seeing performance issues with? I have seen cases where computer resources are quite limited (4gb ram, 5yr old processor, etc.) where vision can be a bit laggy.
Add Hugo in the loop.
We are use a quite new laptop, ThinkPad T14 Gen 2, I7 11th Generation, 16GB memory. The hardware shall be not an issue. We also have been talked with Hogo 2 month ago, it should be related to the network issue. And checked the Laerdal-suz.tulip.co instance is hosted in AWS US, and there are not option in AWS AP.
For those that might run into this same issue in the future, I met with @david.kang and the root cause was 2-fold:
- Intel Realsense cameras require a lot of bandwidth to send their video streams, this means they must be plugged into a USB3.0 port on the computer to get stable video feeds.
- If any USB extenders are used to extend the camera cable length, it must also be USB3.0.
A slick tool to troubleshoot realsense cameras specifically is Intel RealWear Viewer (Intel RealSense SDK 2.0 – Intel RealSense Depth and Tracking cameras). It helped identify the USB port as a 2.1 port.
Hope this helps-