Hacker News Reader

TBH, I think you over-complicate this.

1. Dry-weight scales are very common in produce/stores and extremely common in EU stores (I'm in Russia right now and it's de-facto the standard for most every store)

2. Get a few scales cause you can use them for all produce year around, and they easily earn their keep, and the upkeep is cheap

3. Pre-pick pumpkins under a certain threshold and sell at flat prices; this creates two sale levels: one for those who want a big pumpkin to mess with, one for those who just want a pumpkin to carve. The latter doesn't need to be heavily regulated, your workers can just eye-ball it and given the margins around Halloween you won't really lose out (probably such "eye-balling" will benefit the farm, not the consumer)

The technical solution here might eventually be more accurate, but it doesn't really benefit anyone. The conditions to make the camera detection work are wonky and you probably end up wasting more time making sane conditions for detection than you do just setting some limits.

Similarly, don't forget that too much tech, and it actually might __discourage__ people as they see the farm as non-authentic. The aesthetic of an organic "real" farm doesn't include a lot of high-tech in many people's minds, and you risk offending the wisdom of the masses by being "too technological". Granted, there's absolutely no difference in the end result, but consumers certainly thing differently about it (there's a reason a lot of people still prefer a human cashier to the touch-screen terminals at fast food restaurants, and it's not efficiency, that's for sure)

You could also pre-label any more exotic (type/size) pumpkins.

For that matter, you could pre-label everything outside of a "commodity" which you could segregate. But just weighing seems more straightforward than messing with boxes and seems like it would be a pretty good proxy.

The way they solve this at a huge pumpkin patch out in Snohomish, WA (not Craven) is by having a member of staff walk down the line and tally up before people hit the register. They write the sizes and sum on a piece of paper, you give it to a cashier. Done. Lines went QUICK.

Low tech, works well.

The patch we went to had a loop with a string attached. Loop went over stem and they wound the string around the pumpkin. String had markings so they could tell how much to charge. Took all of 2 seconds per pumpkin.

I don't really want a teched out pumpkin patch.

Weight is the simplest by far. 50,

Otherwise, build a box, and the smallest box that fit the pumpkin determines its size.

I wonder if they run into issues selling by weight. Customers are usually paying more for the visual size, not however much water a particular item has in it.

Does the weight increase linearly with size? If it’s especially hollow, probably not, while an especially large and hollow pumpkin may be uncommon and should be sold for a high price.

Edit: I guess a better question would be, can you predict desirable carving qualities and visual size accurately as a function of pumpkin weight?

FWIW: Selling by weight seems to be common (and quick) in Minnesota.

Take a "sample" of pumpkins, weigh them, have the humans write down their prices and fit a pumpkin weight -> price model. A pumpkin is sort of a hollow shell if you discount the weight of the goop inside, so the weight should increase with the square of the diameter.

Weight is beautifully simple, but the process of unloading 100lbs of pumpkin to be weighted & loading them back in your cart might be a significant driver of the queue latency.

Grocery stores pipeline the unloading & loading steps.

Just roll the wheelbarrow right over an in-floor scale! Subtract the wheelbarrow's own weight, which can be printed on the wheelbarrow itself. Or for something more high tech, put a QR code on the wheelbarrow and you can periodically update its empty weight in the database when it's empty and scan it at checkout.

This is basically how it works at my local landscaping store where they sell gravel by weight.

That wheel barrow is probably half full of water bottles, coats, and whatever other random crap the kids needed. Probably a kid in there too!

I’ve sold pumpkins on a farm before. It’s a social experience where you pay for the pumpkin “ticket” at the end.

Whole families dress up for the experience. People love being in a pumpkin patch and searching for their favorite pumpkin. They love doing it with their family. They enjoy the risk of wheeling them and guessing how much they weigh.

And the farm makes more money when they grow heavier pumpkins.

This isn’t an activity to be taken so seriously.

Scales and measurement aren’t really necessary. I could eyeball and wobble a pumpkin very quickly and give a fair price. But then I am having all the fun, not the customer.

Put the tech into making it a better experience!

Oh is that tractor pulling a wagon? Got to go!

Here is my thought process:

> - Move your barrow under a camera.

Minimum $5k, closer to $25k, to set physical hardware, network, queueing line and instructions

> - It would take a picture and through image recognition identify and classify pumpkins based on size.

$100k+ capex to get implemented model on hardware, sufficient training samples + training expertise, QA, etc. Ongoing opex for maintenance (sun, weather), model maintenance, and cloud infra / use

> - This would be pushed to cashier with total.

Customers would argue with cashier, much like self-checkout lines at big box stores.

> - Maybe where confidence level on classification was low it would prompt human intervention.

Confidence could be high on false positives/negatives.

-----------------------------

Overall, I like the idea of AI-assistive technology implementations. Are we to the point where it won't wreck the economics of a small operation?

@pumpkinman - IMO in regards to business viability it’s just as valuable to see someone’s process of how they came to the conclusion as the conclusion itself

Thanks for the support. I considered the @pumpkinman's complaint quite odd!

To address their primary note directly: I do indeed speak this way when reviewing details and considering hunches on viability. If someone only needs a conclusion then I give them one. Most people legitimately considering something, like creating an ML implementation (something I do all the time) and mapping to business feasibility, benefit by understanding the "why" of a conclusion. That by chance you do not personally benefit is not pertinent. If I have read your comment appropriately, I recommend adopting this interpersonal wisdom: https://xkcd.com/1053/

True that’s a steep one time cost for one patch, but then sell the process and tech to your competitors.

Or get a pumpkin size detector working as a phone app and load it on a cheap $150 android phone taped to a stick

A pumpkin size detector is also known as an eyeball.

Since the demand is concentrated over such a short period I guess the simplest solution would be to have one price for all?

This would encourage people to shop early as they would get the pick of the crop - so to speak. Latecomers to the party would have less choice, obviously, but would still be incentivised to buy anyway since there are not that many alternative pathches to choose from in a given area.

The price could also be reduced as time goes by so the day before halloween you're almost certain to sell the most amount possible.

Pumpkins are an expiring good. To move as many as possible as quickly as possible while also collecting as high of a price as each person is able/willing to pay, it is best for the seller to offer many versions of the pumpkins at many price points.

Aka selling different styles of pumpkin at different price per weight. The seller can sell to people willing to pay more for various styles, and people willing to pay more for various sizes. And to people wanting to spend less, and everyone in between. Which is what already happens.

I would agree with you but, from my limited observation, there seems to be an awful lot of pumkins left over after halloween has been and gone. Maybe what already happens isn't the best way?

I have no idea. I generally assume markets with many participants that have existed for many years and with seemingly no regulations, are the way they are because various constraints of nature have tended towards that solution.

One would need to know what a typical amount of waste is and the cause of that waste to determine if there is too much waste and how to fix it. But I am guessing whatever waste there is now is a pretty optimal amount of waste unless there is some new development that reduces the volatility of producing pumpkins.

Farming in general is like that, since so many variables are outside of peoples’ control. You have to aim for extra even if you are not projecting to sell all of it, assuming the cost of making the extra is absorbable.

Unload barrow on a checkout cage, top of checkout cage is a plank with holes on top with one hole for each size category (the bigger the more expensive)

Pumpkin falls on an Arduino controlled button, adds +1 on a price multiplier

All pumpkins get unloaded on a checkout wheelbarrow for customer to load on their car.

To open the cage and collect the checkout wheelbarrow you have to pay price is sum of multipliers times their respective base price.

Unsold loaded checkout wheelbarrows in cages can be pushed by staff at the beginning of the queue with a "pay with credit card and leave in 30 seconds" sign, at a 10% premium.

If each pumpkin was passed down a conveyor belt in a single file line then it's diameter could be easily measured with an array of light sensors aimed vertically which could be used to set the price in the checkout system. Different customer's orders could be separated by a board that spanned the full width of the sensor array, which would stop the conveyor until it was removed.

Okay. You fund it, I'll build it.