- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi! Please tell me how much time it takes to analyze one image for the maskrcnn model on neural stick 2? My tests showed that it takes about one minute to analyze. Is there any way to reduce the processing time or is this the maximum capability of this device
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
This is expected due to the different between device characteristics.
One of the workaround is Myriad optimization for MyriadX ( NCS2 need to have at least 4 inference requests in parallel).
You can refer here: https://docs.openvinotoolkit.org/latest/openvino_docs_optimization_guide_dldt_optimization_guide.html#myriad
And also here: https://docs.openvinotoolkit.org/latest/openvino_inference_engine_samples_benchmark_app_README.html
Sincerely,
Iffa
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi! Can you tell me what this parameter VPU_HW_STAGES_OPTIMIZATION is responsible for and how it affects the network load speed? It turned out that if the value is YES, the neural network takes a very long time to load (about 10 minutes)
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Greetings,
This is known as a bug and we might expect this to be fixed soon later.
The option VPU_HW_STAGES_OPTIMIZATION exists for debug purposes. When you turn it off, all the hardware acceleration gets disabled for the model. It might affect various characteristics like accuracy, memory requirements, and performance.
Fow now, it is recommended to turn the function off (set it to "NO")
Sincerely,
Iffa
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Greetings,
Intel will no longer monitor this thread since we have provided a solution. If you need any additional information from Intel, please submit a new question.
Sincerely,
Iffa
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page