- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I inference a yolov4-tiny network with an image of 832*832 resolution on one NCS2, and it works well. For some reason, I need a bigger input(image 3328*3328 resolution), one NCS2 makes some error("Not enough memory to allocate intermediate tensors on remote device"). I wonder if I use several NCS2 at the same time to resolve this problem?Thank you!
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
A higher resolution image will still not work even on multiple NCS2.
Using multiple devices speeds up inferencing by parallelizing the processes. As the error message says, there is not enough memory on one NCS2 to support that high of resolution. So even if you are using multiple devices, the image will need to be processed by a single NCS2 and it does not support a resolution that high.
We recommend using the lower resolution image for better results if it is possible.
Redards,
Mauricio R.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thanks for reaching out,
We are looking into your issue and we will come back to you soon.
Regards,
Mauricio R.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
A higher resolution image will still not work even on multiple NCS2.
Using multiple devices speeds up inferencing by parallelizing the processes. As the error message says, there is not enough memory on one NCS2 to support that high of resolution. So even if you are using multiple devices, the image will need to be processed by a single NCS2 and it does not support a resolution that high.
We recommend using the lower resolution image for better results if it is possible.
Redards,
Mauricio R.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page