Pollen_viability

Routine Pollen Detection Manual

This guide walks you through the full process of running an automated pollen viability analysis — from preparing your images to downloading your results. No programming experience is required.

[!NOTE] This pipeline runs the heavy computation on a remote university server (CESNET cluster), so it works even on modest laptops. You only need to start the job from your computer — the actual image analysis happens remotely.

Before You Start: One-Time Setup

You only need to do these steps once when setting up the workflow for the first time.

Step 0.1 — Get Access to the Cloud Storage (S3)

All images and results are stored in the CESNET S3 cloud. You need access before anything else.

👉 Follow the CESNET S3 Connection Guide

This will take you through applying for access, generating your personal keys, and installing Cyberduck (the tool you will use to upload and download files).

Step 0.2 — Set Up the Detection Script on Your Computer

The detection pipeline is run via a script. You need to have the project code and its required tools installed once on the computer you’ll use to launch jobs.

0.2a — Install Required Software

You will need three programs installed:

Program	What it does	Download
Git	Downloads the project code	git-scm.com
Docker Desktop	Packages the detection code to run on the cluster	docker.com/get-started
WSL 2 (Windows only)	Lets Windows run Linux-style scripts	Microsoft Docs

[!IMPORTANT] On Windows: After installing WSL 2, open Docker Desktop → Settings → Resources → WSL Integration and enable it for your Linux distribution. On Linux: After installing Docker, run sudo usermod -aG docker $USER and log out and back in.

0.2b — Download the Project Code

Open a terminal (on Windows: search for “Ubuntu” or “WSL” in the Start menu) and type:

git clone https://github.com/Pollination-Ecology-Group/Pollen_viability.git
cd Pollen_viability

This downloads all the scripts you need into a folder called Pollen_viability.

0.2c — Download the Cluster Access File

You will also need the kubeconfig.yaml file — this is the key that lets your computer talk to the CESNET cluster. Ask the project lead to send you this file, then place it inside the Pollen_viability folder you just downloaded.

Routine Detection — Step by Step

Once setup is complete, running detection every time takes just 3 steps.

Step 1 — Upload Your Images

Using Cyberduck, navigate to the Ostatni/Pollen_viability/detect_images/ folder in the S3 bucket and drag your microscope images (TIF or JPG format) into it.

[!TIP] You can upload a whole folder at once. Just drag it into the Cyberduck window.

Step 2 — Launch the Detection Job

Open a terminal (on Windows: Ubuntu / WSL) and navigate to the project folder:

cd Pollen_viability

Then run:

./deploy_pollen.sh

[!NOTE] The first time you run this, Docker will download several components. This may take 5–15 minutes depending on your internet speed. Subsequent runs are much faster.

You will see output in the terminal that looks like this:

🌸 Pollen Detector Deployment Script
🚀 1. Building Docker image...
☁️ 2. Pushing image...
🚀 4. Deploying Job to Cluster...
👀 6. Streaming logs...

The terminal will then show live progress from the cluster. You don’t need to do anything — just wait.

[!TIP] If you see a permission denied error when running ./deploy_pollen.sh, try:

Linux: sg docker -c ./deploy_pollen.sh

Windows (WSL): Make sure Docker Desktop is running and WSL integration is enabled.

Step 3 — Download Your Results

Once the detection finishes (the terminal shows it has completed), your results are automatically uploaded to the S3 bucket.

Open Cyberduck and navigate to:

Ostatni/Pollen_viability/detected_images/ — Annotated images with coloured bounding boxes showing which pollen grains are viable (green) and non-viable (red).
Ostatni/Pollen_viability/pollen_counting_results/ — CSV summary files (pollen_counts.csv) with counts and viability percentages per image.

Download these files to your computer by right-clicking → Download in Cyberduck.

Troubleshooting

Problem	Solution
`permission denied` when running script	See the tip in Step 2 above
Docker Desktop not starting	Make sure WSL 2 is installed and enabled
Script says “cannot connect to cluster”	Check that `kubeconfig.yaml` is in the `Pollen_viability` folder
No files appear in S3 after job finishes	Wait a few minutes; large jobs can take time to upload results
Any other issue	Contact the project lead or schedule a meeting