Boris / Chinola Zoom Summary

Meeting notes, restricted transcript access, and follow-up actions.

Date: 2026-06-02 Time: 10:35-11:24 EDT Duration: 49m 18s Zoom: 84303573343 Captured by Max's Agent Max Miami, Dmitry, Boris, Maxim Dominicana

Summary

The meeting focused on whether a drone and computer-vision workflow for Chinola/passion fruit can become useful to real farms, and what the earliest measurable experiment should be.

Max framed the current stage as exploration, not yet a committed product. The near-term goal is to test whether AI can detect and count visible fruit, identify fallen fruit, and eventually surface health, disease, flowering, and fruit-set indicators. A low-cost pilot can start with phone photos, roughly 200-500 representative images from different angles, plus any drone or public data.

Boris pushed on product usefulness. A person walking the field with a camera does not solve much for a farm, because a human can already observe many symptoms directly. The more valuable version is likely a drone or automated system, but there are constraints: camera quality, flight automation, viewing angle, and leaves blocking fruit.

A key agronomy point: Chinola fruit stays green on the vine for most of its life. It yellows close to maturity, sometimes only right before falling or once on the ground. Detecting only yellow ripe fruit is not enough; the model likely needs to detect green fruit, flowers, fruit-set stages, and fallen fruit.

Decisions / Direction

  • Treat this as an experiment first, not a finished farm product.
  • Start with the cheapest data collection path: phone photos and existing drone footage/photos before buying hardware.
  • Use detection/counting accuracy on visible fruit as the first useful metric.
  • Track occluded fruit and green fruit as separate failure modes.
  • Drone viability depends on angle, camera quality, and whether side views or planned flight paths can see fruit hidden from top-down views.

Meeting Details

  • Meeting date: 2026-06-02.
  • Meeting time: 10:35:35-11:24:54 EDT. Vexa raw time: 14:35:35-15:24:54 UTC.
  • Meeting URL: https://us06web.zoom.us/j/84303573343
  • Meeting ID: 84303573343.
  • Participants: Max Miami, Dmitry, Boris, Maxim Dominicana.
  • Participant emails: not captured. Needs calendar event/Zoom roster integration or manual attendee email list.
  • Summary email: sent only to Max because attendee emails were not captured.
  • Transcript coverage: captured through 11:22:48 EDT, about 47m 13s into a 49m 18s meeting.
  • Capture gap: final 2m 05s has no transcript segment, so goodbye/closing remarks may be missing.

Next Steps

Max Miami

  • Define the first measurable test: fruit count/detection accuracy on a small labeled set.
  • Prepare a photo collection guide covering angles, distance, lighting, fruit stages, flowers, fallen fruit, and canopy occlusion.
  • Build or reuse a 200-500 image sample from phone, drone, and public sources.
  • Run the current detector and report precision, recall, and obvious failure modes.
  • Keep wording as experiment-first when discussing this with farm owners.

Boris

  • Share representative plantation photos/videos: side views, top-down drone views, flowers, green fruit, yellow/fallen fruit, dense leaf occlusion.
  • Confirm the most valuable farm question: yield count, fallen fruit, disease/health, flowering/fruit set, labor planning, or general monitoring.
  • Share drone model, camera resolution, typical flight altitude, sample footage, and whether side-angle/manual flights are feasible.

Dmitry

  • Help document visible passion-fruit growth stages: flower, fruit set, small green fruit, mature green fruit, yellowing, fallen fruit.
  • Help define labels/classes so the model is not only detecting yellow ripe fruit.

Everyone

  • Agree on one pilot success metric before buying hardware or expanding engineering scope.

Risks / Open Questions

  • Can top-down drone footage see enough fruit, or does Chinola require side-angle capture?
  • Is the current farm drone camera good enough for fruit-stage detection?
  • Which operational question matters most to the farm: yield estimate, fallen fruit, disease, flowering, or general monitoring?
  • How much of the hidden/occluded fruit can be estimated rather than directly detected?

Transcript

Full transcript and English translation are restricted. They should live in Google Drive with explicit attendee access, not on the public wiki.

  • Google Drive URL: https://drive.google.com/drive/folders/1iMDl33ka67s4ZR-f3cpMzcHTkS54hAlv
  • Restricted transcript target: My Drive/Chinola/Meetings/2026-06-02-boris-chinola-zoom/
  • Access policy: Max controls access. Folder is owned by [email protected]; [email protected] has reader access. Attendee emails should be added as readers when known; everyone else must request access.
  • Share controls: Drive permissions should be managed by Max or by the post-meeting script after Google Drive auth is configured.
  • Contains: original Russian transcript, English translation, summary, action items, and capture metadata.