- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I'm an engineer student in electronics and I'm doing an internship about computer vision on autonomous vehicle. I'm currently looking for embbeded solution for computer vision.
My goal is to find a system (board + 1 camera) who can do real-time object detection (by real-time I mean the object detection has to be able to process around 45 FPS, with a resasonable resolution for the images). So I have some questions :
-What embedded product from Intel can handle my goal?
-For now I saw two products : the Neural Compute Stick 2 but with other product, or the Arria® 10 GX FPGA Development Kit who is way mor expensive. What's the better option?
-The software necessary is OpenVino? Is it free or how much it costs?
Sincerely,
Olivier Berton
Stagiaire Expléo Nantes
téléphone personel : 06 81 51 63 04
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Olivier,
You can use any of the development kit from the OpenVINO supported devices.
The Intel® Neural Compute Stick 2 is powered by the Intel® Movidius™ Myriad™ X (VPU) and thus only FP16 model format is supported whereas for Intel® Arria 10 FPGA supports FP32 and FP16 model formats. Also there are other differences in terms of:
- Supported Input Precision
- Supported Output Precision
- Supported Input Layout
- Supported Output Layout
- Supported Layers
So, you can choose any of them as per your requirement.
You can download Openvino and use it free of cost.
Best Regards,
Surya
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Olivier,
Thanks for reaching out.
You can also check these benchmarks for you to make the best decision on which hardware suits the needs for your applications.
Regards,
David C.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page