FlowerChecker: tell me what this plant is | Hacker News Reader

FlowerChecker: tell me what this plant is

blog.flowerchecker.com

90 pointsnkurz23 comments

I love the idea of this. I have a casual interest in gardening and often come across plants I'd love to know the name of.

I hope you go on to great success. Finding niches like this to run a start-up is awesome.

In response to the other comments, perhaps you can make the botanists a 'knowledge marketplace' whereby they are paid (a micropayment of sorts) for sitting through and identifying flowers from pictures. You can set localisation filters so that avid gardeners can get involved too.

As a people that does exactly that since a lot of years just for fun, to be paid for continue identifying plants of the world will be a interesting change, but be aware that they will need to compete with unpaid positions right now. Their main problem will be to make people want to pay for something that they have for free (in an amateur base). Helpmefind for example is still trying to make money with this type of business.

I agree, its hard to monetise something which people do for free and for fun however as someone who considers themselves to be a casual gardener, just remember the audience you are trying to reach is not the hobbyist... its the mainstream masses.

I'm sure some VCs will love this idea. Find the right ones that don't push you to monetise right away. Focus on user acquisition, then monetisation will come later.

This is great, as I find myself far too often squinting at some green thing and wondering WTF it is, BUT....

Having botanists do this is good for accuracy, horrible for scalability. A neater solution would be to, as the botanists identify species, fill in a number of key datum for them, e.g.

Where is it growing?

What colour are the flowers?

How many nodes do the leaves have?

Are the leaves serrated?

Is this plant parasitic?

etc.

Ask enough appropriate questions and you can identify the vast majority of cases - essentially "animal, mineral or vegetable" - thus freeing up the botanists to do the really tricky ones.

Yep, actually, we are going to develop a kind of hybrid between fully automated plant id service and classical botanist key. I made a this task a topic for student's thesis.

Please take in mind that this will be really hard for neophytes. Botanical terms are serial killers even for experts, and some families are a real nightmare. You'll probably lost the 95% of the people in the first 10 serious questions.

You'll need to put thousands (maybe even hundreds) of questions for some species.

> Are the leaves serrated?

You can put this with nice image/grapics of serrated/non-serrated leaves. This idea remind me of Akinator, where you can answer Yes, Proabbly, Don'tKnow, ProbablyNot, No. (http://en.akinator.com/ ) (You also have to consider the possibility of a mistake in one of the answers.)

One problem is that plants have different names in different parts of the word, or even in different parts of the same country. Another problem is that the same name may describe different plants in different places. (Perhaps the Wikipedia article can be the "official" identifier.)

Well, in botany is more a question of: "are the leaves entire, serrulate, serrate, double serrate, dentate, crenulate, crenate, toothed, spiny or undulated?". A lot of people just will walk away when things go a little hard, probably feeling cheated. The problem is that people are just lazy.

I can probably identify thousands of genus plants at least at first sight but is not easy and sometimes is impossible. There are about 3000 spurge species for example (and maybe ten real specialists in the genus Euphorbia in the entire world).

There are also some legal troubles in the field to care for. Some typical recurrent doubts in plant identification in my experience are:

1-"How can I cultivate peyote/cannabis/some ilegal drug in my house?, Is this a male or female plant? Is this cactus in my public park a San Pedro?"

2-"Is this leaf/mushrom/berry that I'm cooking edible or poisonous? (typically with a blurry small photo of some green poor thing that falls between a child in cucumber disguise and an alligator)."

Both 1 and 2 situations could easily escalate to a lawsuit against you. Assure you to put a big disclaimer note in the terms of use.

3-"Urgent!, Urgent!, I want you to do my schoolwork. Identify my entire herbarium, Is for yesterday!."

This guy will not even remember to thank you later most of the times, nor will consider to paid a cent for this (I will gladly charge him for my time if I could otherwise).

This reminds me of this xkcd comic http://xkcd.com/1425/

We've all but nailed the image recognition problem now though.

Suddenly, a leopard print sofa appears, though... [1][2]

    [1] https://news.ycombinator.com/item?id=9749660
    [2] http://rocknrollnerd.github.io/ml/2015/05/27/leopard-sofa.html

We're getting super close! I'm thoroughly impressed with Google Photos new search functions.

Inspires a lot of ideas: Pest Identification - take a picture of an insect infesting your house, have it identified, get a result and maybe a recommendation, serve up pest control businesses in your area (pay a premium to be listed).

Computer vision variations on this: http://leafsnap.com/ and https://itunes.apple.com/en/app/dogsnap/id532468586 (funny at parties: what kind of of dog are your friends?)

I've had a similar project I've wanted to get started on for a while now.

Does anyone know of a good API or Database for plant identification? Something where you can provide leaf # and shape, flower type and color, location, etc... and be returned a list of likely plants would be ideal but a database with that information is definitely workable.

I like databases, and can do things really awesome, but the problem with your focus is that databases do difficult things easier but easy things almost impossible. I'm very confident that I would beat any newbie with a database in most of the cases, even when the plant is totally unknown to me.

I have created a lot of databases to identify some types of organisms of my interest and often find dificult to achieve good identifications only with databases alone.

I also want to build my projects and get money but I'm not sure I want what I call "dependencies" - employees, investors. Probably I can off-load this task to a few closest partners.

As a simple human I can fulfill all my desires and it doesn't cost millions.

Creative journey for me is more important than business race.

I was wondering whether something like this existed for fish. Anyone know?

Is quicker to ask a human but you can also use fishbase. What fish do you want to identify?

man! this is slick!! I love it!!

I was recently thinking it would be sweet if a real-life PokeDex existed and you could identify animals by putting up a phone in front of one and have it be identified.

Hey Arthur, check out http://zoology-app.com/. This is a real life pokedex app currently in the works

Any ETA? I would love to try it on birds. It's a real pain searching for the name of a newly seen bird.

Edit: Aah bummer, seems won't be available for android anytime soon.

Ref: http://zoology-app.com/where-can-i-get-zoology-2/

Reply on news.ycombinator.com