10-20-2020, 06:19 PM
I noted that there's a call to action for people to develop things for the SOEdge, but I also noted that it's hard to determine where its niche is felt to be. That might make it difficult for chickens to provide eggs that others can consume. 
However, I was looking at the upstream Github for the BSP and wasn't immediately finding licensing information or what the difference between the full and Neural AI Stick BSPs is. Does anyone have this information readily available? Downloading 25 GB of BSP SDKs to compare them, for a board I don't have, isn't high on the "this is a fun way to get started" list.
But, more than anything, what do you (both PINE64 and the community) see as the most likely use cases and software to be run on them? That will help interested developers decide if it's a project where their $50-60 + time is going to be well spent by producing output that is going to continue to be useful to a community.
It's obvious that it is positionable for AI, but that's still a very wide range of uses (machine learning, speech, vision, language processing, robotics, expert systems, and more...) and associated software. I see a viable approach in bootstrapping a JeOS baseline image to validate functionality and performance, and then building images based on that for specific tasks.
I know that most buying the board will want to use Acuity with the Vivante NPUIP, but what sorts of things will people be doing with the NPU? What sorts of non-NPU things are people wanting to do?
I'm interested in hearing from the buyers and potential buyers of the SOEdge,of course. What do you want to see this become? Are you planning on basing a business model on using these boards? What do you need to be successful? As a hobbyist, what will make your life easier and save you time by already having common stuff in place?
I'm not explicitly volunteering to start putting some images together, partially because I don't have the hardware to test on right now, and partially because I don't know what would be useful to others. But I expect that anyone else who might want to build images for wider consumption than themselves will also want to know how to make them as useful as possible.
However, I was looking at the upstream Github for the BSP and wasn't immediately finding licensing information or what the difference between the full and Neural AI Stick BSPs is. Does anyone have this information readily available? Downloading 25 GB of BSP SDKs to compare them, for a board I don't have, isn't high on the "this is a fun way to get started" list.
But, more than anything, what do you (both PINE64 and the community) see as the most likely use cases and software to be run on them? That will help interested developers decide if it's a project where their $50-60 + time is going to be well spent by producing output that is going to continue to be useful to a community.
It's obvious that it is positionable for AI, but that's still a very wide range of uses (machine learning, speech, vision, language processing, robotics, expert systems, and more...) and associated software. I see a viable approach in bootstrapping a JeOS baseline image to validate functionality and performance, and then building images based on that for specific tasks.
I know that most buying the board will want to use Acuity with the Vivante NPUIP, but what sorts of things will people be doing with the NPU? What sorts of non-NPU things are people wanting to do?
I'm interested in hearing from the buyers and potential buyers of the SOEdge,of course. What do you want to see this become? Are you planning on basing a business model on using these boards? What do you need to be successful? As a hobbyist, what will make your life easier and save you time by already having common stuff in place?
I'm not explicitly volunteering to start putting some images together, partially because I don't have the hardware to test on right now, and partially because I don't know what would be useful to others. But I expect that anyone else who might want to build images for wider consumption than themselves will also want to know how to make them as useful as possible.