hilum

SeedCounter by Hilum

Count and measure seeds from images.

Load a photo or a scan. SeedCounter proposes the outline of each seed, you check it, and you get the count, viability and measurements in millimeters, with a test report. It opens in the browser, with nothing to install.

Open SeedCounter Why it runs in the browser
  • No install, no sign-up
  • Your image never leaves your computer
  • Works offline after the first visit

real image · 38 seeds annotated by GPEOrq

On seed researchers' benches

  • Prof. Dr. Nelson Barbosa Machado-NetoOrchids and forage species · ISTA and IUCN
  • Prof. Dr. Ceci Castilho CustódioSeed technology · ABRATES
  • GPEOrq and GPSEMThe research groups where the app was born
Who makes SeedCounter

How it works

From photo to test report, with you checking every number

The app does not decide on its own. Every proposal appears as a dashed outline and only becomes data when you accept it.

SeedCounter screen with twelve soybean seeds outlined in cyan on a gray background, a millimeter ruler along the edges and the results panel showing 12 viable seeds.
Real screen captureSeedCounter 4.0 · soybean scene with a known count · 12 seeds, one click on each

  1. Load the image

    A photo from a phone or camera, a scan from a flatbed scanner, a multipage TIFF or the live camera. The ruler in the image gives the scale.

  2. The machine proposes

    One click and the wave outlines the seed out to its edge. For large batches, the optional AI suggests the seeds for the whole queue.

  3. You check

    Accept, fix the outline or change the class: viable, non-viable or the classes in your protocol. Nothing goes into the count until you have checked it.

  4. Export

    Test report as PDF, CSV spreadsheet, SQL database and a dataset for training models. Measurements come out in millimeters, with the scale recorded.

See the step-by-step examplesHow the test report is calculated

Test your eye

How many seeds are on the plate?

The plate appears for a few seconds. Make your guess and watch SeedCounter count them one by one.

More animations in Learn

The plate stays on screen for 4 seconds and then disappears.

Plates generated on the spot, with model seeds. In the app, you count your own image: open SeedCounter.

What it does

Each feature twice: as an idea and on screen

On the left, the animation that explains what happens. On the right, the same step in SeedCounter. Ideas are born as animations and become tools.

One click, one outline

The seed is separated from the background by color and the outline is measured on the spot.

Otsu threshold, outline and Feret

Real recordingone click per seed: 12 outlines, each measured in about a tenth of a second

From plate to histogram

Seeds stack up by length and the seed lot becomes a distribution.

50 model seeds

Real recordingversion 4.0, real soybean scan (Syahraza et al. 2025, CC BY 4.0): the gallery stacks the 20 whole seeds by length, in pixels, because this scan has no ruler

Measured in millimeters, not in more or less

A ruler photographed next to the seeds becomes the scale. The file's DPI is only a declaration.

20 mm = 3,790 px · 1 px = 5.28 µm

Real recordingversion 4.0, real GPEOrq scan with a ruler: dragging from one centimeter mark to the next measures 1,885.5 px, and the scale comes out at 5.304 µm/px

Each seed with its own data sheet

Area, circularity, solidity and the Feret diameters of each seed, with the axes drawn on it.

The numbers of a shape

the rotating caliper

Real recordingversion 4.0, real soybean scan (Syahraza et al. 2025, Mendeley Data, CC BY 4.0): in the Inspector, one of the 20 seeds shows area, solidity, circularity and dimensions in pixels, with the axes on the image

Viable or non-viable, with shape as well as color

Viable and non-viable seeds get marks of different shapes, a disc and a ring, so the figure stays legible even printed in black and white.

tetrazolium, 36 of 65 viable

Real recordingversion 4.0, real image of orchid seeds in tetrazolium (Sementes de Orquídeas, GPEOrq, CC BY 4.0): the analyst marks 36 viable seeds with a cyan disc and 29 non-viable seeds with a magenta ring, as in the group's ground truth, and viability comes out at 55.4%

Germination

Counts day by day and at each withdrawal from storage, comparing treatments and seed lots.

Storage and viability

Queue and batches

Several images in a queue, multipage TIFF and a scan sliced into plates.

For everyday lab work

Live camera

Capture straight from the camera connected to the computer, without going through another program.

The right bench

Saved sessions

Come back to a plate later, with the image, the marks and the outlines.

Open SeedCounter

Who it's for

Choose where to start

The app opens straight to the workbench. If you like, pick a profile on the opening card: it only adjusts what appears on screen, takes effect right away and can be changed later.

Seed testing lab

Many plates a day: count, issue the test report and move on to the next one.

Test reportQueueExport
For labs

Forage species

Full or empty spikelet, dormant or non-viable, and germination every week.

ClassesGerminationStorage
Full or empty forage seeds

Learners

Mark, count and see the number, with the app pointing out the next step.

MarkCountTimer
Learn with animations

Classes and presentations

Project only the image and the marks, without changing anything on the computer.

ProjectMarksTotals
Open in presentation mode
Real recordingversion 4.0: the opening card is optional; with Student in training, the examples and the Germination tab go away and the usage instructions appear
Open SeedCounter See usage examples

Examples

From a grain of rice to an orchid

The examples are being made with real seeds: photographed at the bench, analyzed in SeedCounter and shown step by step, with the final result.

See all the examplesThe app's demos

Science

Measured against the literature, with the errors in plain view

Before it becomes a feature, each measurement is checked against published data. What we still get wrong is written down.

7 of 11

Descriptors identical to the published dataset

On 1,904 rice grains measured by SeedCounter and in the Koklu et al. dataset, seven descriptors matched with 0.00% error. The other differences come from calculation conventions.

The rice example
1,0548

The perimeter constant, derived and checked

Theory predicts 8(√2−1)/π = 1.054786. The published dataset shows 1.054052. A difference of 0.07%.

Shape and geometry
+9,9%

What we still get wrong

The outline simplified to 48 sides overestimates circularity in seeds with low solidity. That is measured and explained, and the fix is the next step.

The theory behind the measurements

Blog

Seeds, images and measurement

One idea a week, with the animation, the math and what we still get wrong.

The first post is coming soon.

See all posts

For labs

Light decides the outline. The bench decides the light.

With raking light, a seed's shadow touches its neighbor and two become one. If you're going to use it every day, we work with you on what goes around the app: the capture bench, the classes in your protocol and validation with your own samples.

SeedCounter in your lab Write to us

Where it came from

Born at GPEOrq, counting orchids

SeedCounter began as an answer to a bench problem at GPEOrq and GPSEM, the orchid and seed research groups: how to count and measure orchid seeds, among the smallest there are, without spending the whole day at the microscope. The group had been reading tetrazolium tests from scanner images since 2011; the app turned that reading into counts, measurements and a test report.

Read the storyThe group's papers
The group, in green shirts, poses in two rows in front of the 23rd Brazilian Seed Congress backdrop.
Real photoGPEOrq and GPSEM at the 23rd Brazilian Seed Congress, Foz do Iguaçu, August 2026

Open it, load an example and count.

The app comes with example scenes of known count, so you can compare what it finds.

Open SeedCounter