OpenCores

OpenCores

Neural networks on VHDL

First

Prev

2/2

no use

no use

RE: Neural networks on VHDL by mike_from_space on Apr 7, 2019			mike_from_space Posts: 19 Joined: Sep 11, 2018 Last seen: Apr 11, 2021
Hello, I have been reading books and papers to get information about what I am trying to do all this time. My question is can I see a significant difference of training time when I "go" from a Raspberry Pi 3 B+ to a Virtex 5 implementation? Hello, I have been reading books and papers to get information about what I am trying to do all this time. My question is can I see a significant difference of training time when I "go" from a Raspberry Pi 3 B+ to a Virtex 5 implementation?

RE: Neural networks on VHDL by mike_from_space on Jun 18, 2019			mike_from_space Posts: 19 Joined: Sep 11, 2018 Last seen: Apr 11, 2021
I have been reading a lot of neural networks theory all these months. Now I am going to implement it. What I want to ask is how do I connect the OV7670 camera to the Xilinx SPARTAN 3E-Starter?? Thank you. I have been reading a lot of neural networks theory all these months. Now I am going to implement it. What I want to ask is how do I connect the OV7670 camera to the Xilinx SPARTAN 3E-Starter?? Thank you.

RE: Neural networks on VHDL by mike_from_space on Jun 18, 2019			mike_from_space Posts: 19 Joined: Sep 11, 2018 Last seen: Apr 11, 2021
I am following this: http://hamsterworks.co.nz/mediawiki/index.php/OV7670_camera I am following this: http://hamsterworks.co.nz/mediawiki/index.php/OV7670_camera

RE: Neural networks on VHDL by krbonne on Jun 18, 2019			krbonne Posts: 1 Joined: May 5, 2016 Last seen: Jun 18, 2019
Hi Mike, [q]Hello, I have build the neural network and is trained with caucasian people and it outputs the string: "man" or "woman". When I use chinese people, again it is trained on this new dataset and outputs the string "man" or "woman". So it is dynamic neural network. I cannot make it decide in time less than 30 sec. That is why I am trying building it on FPGA. Thank you[/q] Just out of interest. What are the typical memory requirements for an FPGA to implemented a Neural network? I am currently playing around with the Sipeed M1w dock, which is a small development board with a Kendryte K210 SOC. That soc has an buildin Neutral Network core with 2 MByte of RAM, so that's not little. So I have been wondering, if you would want to implement a neural network core in a FPGA, would you not need quite a bit of RAM? And also, could one implement this with external DRAM, or would this require internal block-RAM on the FPGA itself? Do you have any experience with this? What FPGA did you use? What kind of neural network did you implement? How large was the network? How much RAM did you need to implement your network? Cheerio! Kr. Bonne. Hi Mike, Hello, I have build the neural network and is trained with caucasian people and it outputs the string: "man" or "woman". When I use chinese people, again it is trained on this new dataset and outputs the string "man" or "woman". So it is dynamic neural network. I cannot make it decide in time less than 30 sec. That is why I am trying building it on FPGA. Thank you Just out of interest. What are the typical memory requirements for an FPGA to implemented a Neural network? I am currently playing around with the Sipeed M1w dock, which is a small development board with a Kendryte K210 SOC. That soc has an buildin Neutral Network core with 2 MByte of RAM, so that's not little. So I have been wondering, if you would want to implement a neural network core in a FPGA, would you not need quite a bit of RAM? And also, could one implement this with external DRAM, or would this require internal block-RAM on the FPGA itself? Do you have any experience with this? What FPGA did you use? What kind of neural network did you implement? How large was the network? How much RAM did you need to implement your network? Cheerio! Kr. Bonne.

RE: Neural networks on VHDL by mike_from_space on Jun 19, 2019			mike_from_space Posts: 19 Joined: Sep 11, 2018 Last seen: Apr 11, 2021
Hello Kr.Bonne If you have time take a look here: https://arxiv.org/abs/1806.01683 The RAM on FPGA is a problem, that is why they use external memory in the experiments. And also, as far as I am concerned I use a typical 640x480 camera. Also it matters how many hidden networks you have, because the more hidden networks you want, the more cells you need from the FPGA. I am trying to build CNN, which is the network used for image processing. Other networks, such as RNN are used for forecasting, for example. Now I am trying to use the SPARTAN 3E because it has ready interfaces. Thank you. Hello Kr.Bonne If you have time take a look here: https://arxiv.org/abs/1806.01683 The RAM on FPGA is a problem, that is why they use external memory in the experiments. And also, as far as I am concerned I use a typical 640x480 camera. Also it matters how many hidden networks you have, because the more hidden networks you want, the more cells you need from the FPGA. I am trying to build CNN, which is the network used for image processing. Other networks, such as RNN are used for forecasting, for example. Now I am trying to use the SPARTAN 3E because it has ready interfaces. Thank you.

First

Prev

2/2

no use

no use

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.