ωJump

Find the cells · Janelia OpenOrganelle FIB-SEM · · build 2026-09-04 18:19

not signed in

These volumes have no published list of cells. This is the map being drawn.

Loading…

Loading…

Explore — where is there a nucleus?
Ctrl+click each nucleus you can see.
A plain click does nothing — Neuroglancer places annotations on Ctrl+click, and there is no setting that changes it. The point tool is already selected for you, so Ctrl+click is all it takes. The blue box is the region you are being asked about; the yellow one is the whole dataset. Then copy the address bar and paste it below.

None yet.

Filter and show

Everything here filters the nuclei this community has found in the selected volume — there is no published cell table for any of these datasets, so this is the table, and it is as complete as the map above says it is.

within µm of a

Each box has its own “Colab notebook” button. It writes an .ipynb you upload to Google Colab and run to download that box’s EM imagery, and its segmentation on the volumes that publish one. Nothing is downloaded from this page — the notebook does the fetching, and it prints what it found in the store before it fetches anything.

Every other tool in this family classifies a nucleus table somebody else built. The OpenOrganelle volumes publish no soma list and no cell types, and no detector is being written for them — so the table does not exist until the community makes it. You are being sent to the box furthest from any box already visited, which is why reporting an empty one matters as much as finding a nucleus: an unreported empty box is one the sampler will hand out again.

There are two jobs here and they have opposite rules. Explore hands out boxes at random from among the ones furthest from anywhere already visited, because two people searching the same tissue is wasted effort. Identify hands everyone the same nucleus in the same order, because two people naming the same nucleus is the entire point — one opinion is a guess until somebody independent agrees with it.

Identify also lets you pick a set and walk it in order — every nucleus somebody called a lamellar cell, say — forwards and backwards, which is what you want when you are checking a decision rather than making one. A nucleus whose top two labels are tied is filed under neither: the crowd has not picked a winner, and a menu is no place to pick one for them. Ties get their own list.

That word independent is why Identify hides the existing votes until you have answered. A tally shown beforehand turns an opinion into a confirmation, and a consensus built out of confirmations measures how convincing the first answer was rather than what the tissue is. You can still look — it is one click — and the click is recorded next to your answer so the effect of looking can be measured rather than assumed away.

61 volumes are open, listed smallest first with the number of 40 µm boxes each one takes — 1 for the smallest, 3,672 for the largest, 13,031 across the roster. The map above sizes itself to whichever volume you pick and keeps one scale across its panels, so a distance in one is the same distance in another.

Nothing you name here is scored. There is no ground truth for cell identity in these volumes — only what the community converges on. Organelles are a different matter: 114 crops carry expert voxel labels, and a scored organelle trainer built on those is the natural companion to this, not part of it.