ReefCloud Updates
We've been busy! Find a full rundown of latest releases, improvements and fixes to ReefCloud
New feature added - Desktop upload tool
The Desktop Upload Tool enables uploading of Surveys and Imagery into ReefCloud in a faster more efficient way, providing a streamlined workflow for batch uploading large datasets directly from your local machine, including PEARL drive.
Find the Desktop Upload Tool in the top right of the Surveys page with the ReefCloud Portal, next to Create Survey and Bulk Import buttons.
This tool is currently available only to AIMS users with an AIMS email address (ending in @aims.gov.au) as part of initial testing and feedback. Access will be broadened to all ReefCloud users in a future release.
Improved
Improved Image and Point tagging by updating the Tagging Dialog to require users to explicitly select a custom tag when typing in the custom tag search input, preventing accidental form submission without a valid tag selection. This update also includes several small usability enhancements to make the tagging workflow clearer and more robust.
Minor features added & bug fixed
Minor updates to the ReefCloud Portal.
Minor improvements & bug fixes
Minor updates to the ReefCloud Portal.
Improved
For Community Dashboard linked projects, the 'Community Dashboard Region Management' page now allows for latitude and longitude coordinate entry and editing.
Improved text for user guidance within pages.
Implemented 'point walking' into the QC mode of the Classifier to optimise the speed of performing QC checks.
New Feature-vector Model Deployment
We’ve updated the ReefCloud feature‑vector generation process, now live across the platform.
A feature‑vector model is basically a translator between your image and the ML system, pulling out important visual information and turning it into a neat list of numbers the model can understand.
What's changed?
This update was primarily a security and infrastructure upgrade, designed to make our feature-vector generation more stable, efficient, and cost-effective behind the scenes.
It also brings significant speed improvements when generating feature-vectors for new images. The same tasks previously taking up to an hour are now typically completed in a minute!
To ensure everything remained consistent, we ran extensive testing to confirm that these new feature-vectors do not change the classification results in a meaningful way. The analysis showed that existing classifiers behave the same, with no negative impact on model outputs.
Find the detailed analysis report on the ReefCloud feature vector model update here
What to expect?
Same classification outcomes: No noticeable changes to accuracy or class separation - your results should look the same as before.
More secure & efficient systems: Improved backend reliability, reduced compute costs, and a smoother long-term upgrade path.
Only new images use the updated feature-vectors: Previously-processed images keep their old feature-vectors and newly uploaded images will use the updated system automatically.
Do I need to do anything?
No action is required.
The updated feature-vector system is fully deployed, and all existing projects continue to work exactly as before. There is no impact on your historical analyses, and no changes are needed on your side.
If you have questions or need support, please reach out at [email protected].
Thanks for being part of ReefCloud!
Minor update
Minor update to improve consistency of styling and functionality of current features.
The first release of 2026!
Classifier user interface refresh and data export improvements.
Classifier & Image Library
Improved layout to both the Classifier & Image library
Filter panel now on left
Filter selections now visible top of screen
Buttons reorganised
Enhanced visibility of the number of points selected
Pagination of images and patches at bottom of Classifier and Image Library pages (e.g., Image 1 / 9697 | 1-50 of 9697 images)
Train vs Test points in exports
Point classification export file now contains a new column model_data indicating if points were used to 'test' or 'train' the ML model
Filter exports for Country Region, Local Region and Reef Name
More user choice in filtering capabilities when exporting data to make it easier to find the information needed (especially for large projects)
November update
Classifier improvements to enhance user experience.
User guided selection of images to annotate
Choose to filter for all (default), every 5th or every 10th image to annotate in the Classifier, spreading your annotations more evenly across surveys and your project to optimally train the model.
Image quality enhancement
Enhance image quality while in the Classifier
Can be toggled on/off
A presentation-level feature, independent of the ML model
Last updated
Was this helpful?