I’m wondering what people think about ways to begin sharing data. Somebody please correct me if I’m wrong, but my understanding is that, so far, no one has published any data from growing plants in a food computer. There’s quite a bit about building food computers, but I haven’t seen recipes, time lapse plant images, or logs of measurements. [edit 1: see https://twitter.com/kikai_tomato, thanks for the link @spaghet; edit 2: there are a couple greens recipes at https://github.com/OpenAgInitiative/openag_recipe_bag]
I’ve been thinking about this for a while. In the spirit of Cunningham’s Law, it seems the door is wide open for somebody to build a prototype data sharing architecture for others to discuss and improve upon.
Part of the data sharing problem seems to be that, for many people, building a food computer according to the PFC2 design is prohibitively complex and expensive. But, I don’t see why sharing data or recipes has to be tied specifically to the PFC2 design.
@Caleb, my understanding is that your vision for OpenAg is to have many Food Computer and Food Server designs with sharing and cross-pollination of tools, techniques, ideas, and data. Is that right?
A simple idea on how to begin
As a starting point, it would be great for people who want to share data to just describe what they did–any way they can–in as much detail as they have the patience to manage.
If somebody has CSV data from a PFC1, they could:
- Put it in a gist on GitHub and post the link in a forum thread.
- Put it in a publicly shared google sheet.
The obvious shortcoming here is, how do you associate plant images with the measurements? Also, based on this forum thread and this forum thread, I’m not sure if it’s possible to export data from a PFC1.
A more involved idea
Here’s an idea that somebody could build. This might be implementable on a PFC2, but I’m thinking about it more broadly–maybe this would work for any small DIY or commercial hydroponics system:
- Make a twitter account for publishing your data
- Prepare for your grow by ordering seeds and nutrients
- Tweet where you got the seeds, what variety they are, and a picture of the seed packet. For example:
#openagdata #growNum01 #seeds Johnny's Selected Seeds, Premium Greens Mix, part no. 650
- Tweet the manufacturer and name of the nutrients you’re using along with a picture of their packaging. For example:
#openagdata #growNum01 #nutrients General Hydroponics, FloraDuo A and FloraDuo B
- Use a twitter-bot running on a Raspberry Pi to tweet at 30 minute intervals with a webcam picture of the grow chamber and text like this:
#openagdata #growNum01 #log 19:30 22Feb2017 - pH: 6.5, EC: 900, h2oT: 65F, airT: 73F, airH: 64%rh, CO2: 1000ppm
Rationale for sharing with a tweet-bot
The main idea behind my twitter example is that the easier it is for people to post their data, the more data will get posted sooner rather than later, and the quicker structure will emerge around what ways of sharing data are useful and not.
Using a twitter bot to share data would make it easy for people to get started building an open dataset because:
- You don’t have to get a bunch of people to agree on anything up front
- Your Raspberry Pi can be on a private LAN with a dynamic IP and NAT–it doesn’t need a public IP address, domain name, SSL cert, etc.
- Nobody has to build or host a central clearinghouse for sharing plant data
- Using hashtags like
#growNum01as a serial number for the experiment,
#nutrientsfor metadata, and
#logfor measurements ought to give enough structure for a data aggregation bot to follow the feed and make sense of what was being posted.
I’m assuming that with all the progress in machine learning and computer vision these days, it doesn’t make sense to worry about trying to make people publish their data in a highly structured format. Rather, it’s better to aim for lots of data–including images–and put the burden of aggregation on a bot that follows
#openagdata or specific feeds.
Based on what I see on the forum–due to the global nature of the OpenAg Initiative–people are going to be using different seeds, building with different materials, feeding with different nutrients, and operating in very different ambient temperature and humidity. Trying to anticipate and accommodate all that up front is probably more hassle than it’s worth–my guess is that it would be better to plan for organizing loosely structured data with the expectation that data formats will evolve.
@gordonb, any thoughts on this?