- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I am currently working on tensorflow object detection models.
I downloaded SSD MobileNet V1 COCO* from
Then I used model optimizer to convert it with its supported json file, pipeline config etc. Everything is just fine until this step. I got xml and bin files.
I send these files to my Upcore Plus. While I am running python- sample object detection I faced with low FPS on MYRIAD. When I choose -d CPU I got about 1.2 FPS. However in -d MYRIAD I got 0.26 FPS.
I did everything in your documentations. What I missed here?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Semta,
This should be within expectation as the CPU is more powerful compared to the Neural Compute Stick 2.
Regards,
Rizal
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Rizal_Intal,
Thank you for your answering.
I am beginner in these fields. I do not have a roadmap. Do you have any suggestion for this problem. I need around 10 FPS on Upcore Plus X. Should I use different model, library or optimizer? Should I focus on CPU optimization?
Regards,
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Semta,
The easiest would be to change the model to something else that is less compute heavy.
The popular yolo model (some versions are available as Intel pretrained model) is one of the alternative for a fast model.
Regards,
Rizal
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Semta,
This thread will no longer be monitored since we have provided a suggestion. If you need any additional information from Intel, please submit a new question.
Regards,
Munesh
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page