Shelf-rail cameras that show mushroom pickers which caps are ready right now
TL;DR Capwise mounts a camera on a rail above each mushroom shelf. It scans the beds several times a day, marks which button mushrooms have hit grade size and tells the picking crew where to go next. Big farms pay a monthly fee per room for better grades and a yield forecast they can trust.
India grows about 0.28 million tonnes of mushrooms a year, and button mushrooms make up about 73% of that[1]. Button mushrooms grow fast.
A cap that's the right size in the morning can open by evening and drop a grade, so pickers walk every bed several times a day and judge each cap by eye. New pickers get it wrong and good pickers are hard to keep. Abroad it's even worse - 60% of Canadian mushroom farms couldn't hire all the pickers they needed in 2022[2]. I found no public figure for picking wages or labour share on Indian farms, so that's the first thing you need to pin down.
The seed pitched AI robots that harvest specialty crops, and I've reframed it. Full picking robots exist, but they were built for Western wages and I couldn't find a price that would let one pay back on an Indian farm. So Capwise starts with the robot's eyes and leaves the hands to people.
A camera carriage runs along each shelf tier on a light rail. Every few hours it photographs the bed, measures each cap and colours a room map - ready now, ready next scan, past grade. Pickers carry a cheap tablet or follow a light strip that points to the next bed. The farm manager gets a flush-by-flush yield forecast and a count of caps that opened before anyone reached them, which is lost money made visible.
Pick one crop and one shelving style first, meaning button mushrooms on standard metal shelving in large commercial rooms. A picking arm can come later, once you have a year of images and a farm willing to share the cost.
The owner of a commercial button-mushroom unit with several climate-controlled rooms pays, and these units often sell to retail chains or processors. The picking supervisor is the one who uses it. Nobody publishes a price for picking-guidance tools in India, so my assumption is ₹8,000–15,000 a month per growing room, paid back by fewer caps lost to opening and less time spent training new pickers.
The robot makers have already proved that machine vision can judge a mushroom. 4AG Robotics raised US$40 million in July 2025 and had sold 53 picking robots across Canada, Ireland and Australia[2], with deposits for more than 40 more[3]. Camera modules and edge chips that run small vision models are now cheap enough to put one carriage in every room.
There's no public figure for spending on mushroom-farm automation in India. The bottom-up anchor is the crop itself - about 0.28 million tonnes a year, roughly 73% of it button mushroom[1], grown mainly in Himachal Pradesh, Maharashtra and the hills of Tamil Nadu[1]. How many farms have climate-controlled rooms isn't published, so count them in one region before you size anything.
4AG Robotics sells a vision-guided robot that picks, trims and packs, fitted to existing shelving[4]. It isn't sold in India and its price isn't public. Indian mushroom farms pick by hand, and I found no Indian company selling pick guidance or yield forecasting to them. That leaves room for a low-cost step a large Indian farm can actually pay for this year.
| Difficulty | Moderate (47/100) |
| Starting capital | ₹0.54–1 Cr |
| Time to launch | 9–12 months |
Seeing a cap is the easy bit. Counting caps in clusters where they touch and hide each other, in humid rooms where lenses fog, is where it gets messy. You'll also need farm owners to change how pickers work, and that's a management problem more than a software one.
No sector licence covers a camera system for farms. The rules that apply are about the hardware and worker data.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't trust the model in fogged or poorly lit rooms without spot checks, and don't let it make claims about food safety or residue.
Want to work out the AI layer for Capwise? Talk to Yash →
This is a niche, practical idea and I like it for a founder with computer-vision skills and patience for farm visits. If you're set on building a picking robot from day one then it's a poor fit. First prove one number on a working farm - at least 20% fewer caps lost to opening across two flushes.
Plan for roughly ₹0.54–1 Cr, mostly on hardware prototypes, the vision team and months of farm trials.
No sector licence. You need WPC self-declaration for radio parts, a BIS check on adaptors and cameras, and DPDP compliance if you track pickers.
Yes, large mushroom farms in many countries face the same picking-labour problem. Check local radio, product safety and worker data rules first.
The free Capwise validation kit has the line-by-line capital breakdown, the effort scoring, the radio and data checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for farm owners and supervisors, and 5 assumptions to test. Download the kit →
Thinking about building Capwise or something close to it? The make-or-break work is getting accurate cap counts in crowded, humid beds. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.