Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
tensorflow and V5
#11
(12-10-2018, 02:02 AM)jonsmirl Wrote: It is a HAAR feature classifier. That is an old system from before the development of tensorflow type systems. So it is not clear if tensorflow can make use of the EVE hardware, it depends on how they implemented HAAR, is it microcode or a ROM? Only an AW engineer will know the answer.  It does support customizable classifiers so it may be microcode.

On the other hand, a HAAR classifier is fine for our use case. But we need to be able to retrain it.

Having reviewed the datasheet and kernel drivers, EVE seem to support a bunch of image processing routines along with a DMA engine. There are 3 primary modules in EVE:
  • DMA
  • SF - extracting Pyramid features
  • BOOST -  running Cascade detectors (HAAR/ LBP/ AdaBoost)
Interestingly SF and BOOST can't be run at the same time.

There 64 registers that can be read/written too (unfortunately no documentation). It might be possible to reverse engineer this by sniffing the read/writes by invoking sample EVE applications. Unfortunately I don't have a board to do this.
Reply
#12
There are many parallels between the EVE stuff and OpenCV HAAR support. Once AW releases their converter tool it should become possible to modify the classifiers.

http://www.xavigimenez.net/blog/2010/02/...th-opencv/
https://docs.opencv.org/3.3.0/d7/d8b/tut...ction.html

There are even similar XML files in OpenCV.
https://github.com/opencv/opencv/tree/ma...arcascades

The EVE hardware is probably something similar to this
http://citeseerx.ist.psu.edu/viewdoc/dow...1&type=pdf
This hardware is very different than the computationally based deep learning algorithms.

Visualization of what the hardware is doing
https://www.youtube.com/watch?v=hPCTwxF0qf4
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)