Image Analysis Compute
The parts of image analysis that need a neural network consume compute usage. Everything else, including all the tuning you do before you commit to a run, happens in your browser and is free.
What Uses Compute
| Work | Usage |
|---|---|
| Detecting fluorescent cells, nuclei, or plate colonies | 1 per image |
Detection is counted per image, which makes the arithmetic easy: a run across 40 images uses 40.
A z-stack is counted by slice, not by image. Running a 40-slice stack uses 40 unless you trim it. This is the single biggest lever on what image analysis uses, which is why Conspecta checks focus and offers to run only the slices that are actually in focus.
What's Always Free
Everything that runs on your own machine:
- Uploading, viewing, enhancing, cropping, and annotating images
- The guided builder, and the whole analysis before you press run
- Brightness-cutoff detection, which is arithmetic rather than a model
- Tuning any setting and re-running the preview on one image
- Filtering results by size, confidence, or brightness after a run
- The z-stack focus check, which is scored on your device
- Z-projections
- Every export, of any kind, from any workspace
Reopening a finished analysis is free too. Conspecta restores the detections from the run you already made rather than submitting a new job, so revisiting a result from three months ago uses nothing.
Tip: Tune on one image until the preview is right, then scale. The preview re-runs live as you change settings and uses nothing on a threshold-based pipeline, so there's no reason to run 200 images to find out your threshold was wrong.
Before You Run
Conspecta tells you what a batch involves before you start it. The run button names how many images or slices it will process, and gives a time estimate, since an AI detector takes a minute or two per image and a large batch is a real wait. Turn on the notification bell and your browser will tell you when it finishes.
If a focus check has trimmed a stack, the button says so, such as running 12 of 40 slices, and the estimate reflects the smaller job.
Plans
Detection is available on every plan, but free accounts have no compute allowance, so there's nothing for it to draw on. Settings says so where the allowance would otherwise be.
Tracking It
Settings → Usage shows your team's compute usage as a Subscription allowance that resets each period and any Additional usage the team holds, which never expires and is only drawn on once the subscription allowance runs out.
Below that, Compute Jobs lists every job with who ran it, in which project, what type, when, how long it took, and what it used. Exports appear there with a usage of zero. Export CSV takes the whole history, which is how you work out where a month went.
Usage is pooled at the team level, so every project and every member draws from the same allowance.