Datasets:
Commercial licence request for Oak-Defect-Detection
Hi Nick,
I'm a final-year student at Epitech (France), working on BEAVER, our end-of-studies innovation project (EIP). It's a computer-vision system that grades sawn oak and beech boards against EN 975-1, the European appearance-grading standard: a camera images each board on a conveyor, a detector finds the defects, and a rules engine assigns the grade the standard prescribes.
Your Oak-Defect-Detection dataset is the only public set I've found of green rough oak imaged in a production setting. That makes it uniquely relevant, the other public wood-defect datasets use softwood defect vocabularies that don't map onto EN 975-1 at all. Your Black_Rot and Heartwood masks in particular correspond to criteria the standard uses directly.
As a student project it isn't sold, so I believe our current use would fall within CC BY-NC. But an EIP is meant to become a product, and I'd rather ask now than discover later that a model we shipped was pre-trained on data we weren't licensed to use commercially. So: would you consider granting a licence for commercial use?
What I'd do with it: pre-train a defect-detection model, then fine-tune on our own footage. The dataset itself would not be redistributed. I'm happy to:
credit you and the dataset in the product documentation and in any published model card,
state clearly that the data was modified (converted to bounding boxes, filtered by class),
accept any conditions you'd want to attach, or a paid licence if you'd prefer.
If commercial use isn't something you're open to, that's completely understandable, a clear "no" is genuinely useful, since it lets me stop planning around it.
Either way, thank you for publishing it. It's the only dataset of its kind I could find.
Best regards,
Killian Cottrelle
Hi Killian,
Thank you for reaching out and explaining the BEAVER project. It is encouraging to hear that my dataset is useful for another hardwood computer-vision application, and I appreciate you raising the commercial licensing question early.
I lead computer vision and AI development at Ten Oaks, LLC, a hardwood flooring manufacturer. I created and published this dataset as part of my Fullstack Academy AI/ML capstone project. My work has since expanded to larger, more refined datasets and segmentation models currently used in production.
I am open to considering a commercial license. My primary interest would be a mutually beneficial collaboration, particularly sharing annotated hardwood imagery and technical findings that could support both your project and our model development at Ten Oaks.
Before discussing specific terms, I would like to understand a little more about your plans:
- Is BEAVER currently an independent student project, or are any companies, sawmills, or equipment manufacturers involved? Would the license be held by your team or a company?
- How do you expect to commercialize the system—for example, as equipment, licensed software, or a service? Would customers receive the trained model?
- What imagery and annotations are you collecting, particularly for oak? Would you be willing and authorized to share an agreed portion of that dataset?
- Would you be open to allowing Ten Oaks to use the shared data to train and deploy models in our own production operations? I would also be interested in discussing access to trained model weights or technical findings, where feasible.
I understand that imagery collected at participating facilities may come with restrictions, so it would be helpful to know what you would have permission to share.
If our goals align, we can put together a written agreement defining the permitted commercial uses and what each side would contribute. At this stage, I am open to exploring an arrangement, with commercial permission subject to agreeing on those terms.
If you would like to continue this discussion via email, feel free to reach out at nicholas.rodgers@tenoaksflooring.com
Thank you again for your interest. I look forward to hearing more about BEAVER.
Best regards,
Nick Rodgers
Ten Oaks, LLC