I was wondering if there is any update on when this might be added? The guys from intel do not seem to have a roadmap for it but keep stating 'they are aware of this feature request'. Surely after over a year of requests they can now state if they have actually worked on it, if they are working on it, or if they are not? Rather than giving vague answers can you please state yes or no if this is on the roadmap and if it is when it should be ready? There are lots of people here who need to know so they can make choices about what hardware to use.
Please do not mention about caffe (really not sure why you keep stating this is virtually an equivalent to TF MBN SSD when it is shown time and time again to be slower and less accurate) or Yolo etc. Please just answer the question:
- Is TensorFlow MobileNet SSD on the roadmap to support and if so when should it be released?
Again can @Tome_at_Intel @neal_at_intel or @AshwinVijayakumar answer:
It appears my comment has been deleted so I'll try again.
Caffe is old, slow, and not up to modern uses.
I still do not understand the lack of support for TensorFlow MBN SSD….TF is virtually the defacto environment for all AI models so why is it not supported? This is like building a car and then going….oh there's no seats….and when asked if seats will be added to the car "we may add seats the car, we are aware everyone wants seats in the car but we can't tell you if and when we will add them'.
So I ask again:
@Tome_at_Intel @neal_at_intel @AshwinVijayakumar
** - Is TensorFlow MobileNet SSD on the roadmap to support and if so when should it be released?**
We have tried. We have trained our network with the SSD guide from the two recommended repos:
As well as from other recommended repo sources.
The results are lackluster.
Please do not reply to this very valid question with the words "Caffe" in your response.
It's nice that @Tome_at_Intel @neal_at_intel @AshwinVijayakumar ignore their users and refuse to answer a simple question they should easily have the answer to. I assume your silence is your stating you have no plans for it. As such I think I (and by the sounds, a lot of other people) will be leaving Movidius and i hope this thread serves as a warning to potential people DO NOT USE MOVIDIUS IF YOU WANT TO USE TENSORFLOW MOBILENET!