We hate data annotation. You probably do too. But training AI models requires it, and biology rarely has off-the-shelf datasets. It's grunt work and you could do it yourself or delegate to someone else. Either way, it sucks, it's necessary, and we'll do it for you.

Data annotation frustration

How This Works

  1. You send us a link - Provide us a sharable link to your biological imaging data and describe what you want annotated. The dataset does NOT have to be structured. This is where your work stops
  2. We annotate a sample - We download the data, figure out its structure, do some research, annotate a small portion, and send you samples for review
  3. We get your approval - You approve our annotations or provide feedback. We're not experts in your domain, but we want to get this right. Once approved, we'll provide total cost estimates
  4. We complete annotation - We annotate the entire dataset and notify you when complete. You sit back and relax
  5. You pay us - Pay the agreed upon fair price for the full annotations. We will send you a digital invoice with a link to pay
  6. We deliver your data - We upload the annotations back to the same shared folder. If you provided a reference annotation, everything will be in that format

None of this is automated! A real person will send you an email!

Get Started

Some Annotations We Can Do

Bounding Box Annotation Example

Object Detection

Bounding box annotations for identifying and localizing objects in your images. Perfect for training detection models and object recognition systems.

Segmentation Annotation Example

Segmentation

Pixel-level annotations for precise object boundaries. Ideal for instance segmentation, semantic segmentation, and detailed analysis of biological structures.

Who Is Annotating Your Data?

Microscopy We Work With

Confocal Microscopy Example

Confocal Microscopy

We accept any format of images (tif, png, jpeg, czi, lif, etc...) and of any dimension.

Electron Microscopy Example

Electron Microscopy

Scanning electron microscopy, FIB-SEM, TEM, or others.

Our Annotation Process

  1. Download - We download your data from the link your provided
  2. Organize - Your data doesn't need to have much structure; we'll sort through it and figure out your system. We'll confirm with you that we have a good grasp on its structure
  3. Blinding - We blind ourselves to the data to ensure unbiased annotation
  4. Annotate - Each image is randomly selected and annotated until the entire dataset is complete
  5. Reformat - If you provide a sample annotation, we will ensure the output annotations are in a format you can use
  6. Delivery - Data is un-blinded and returned to you in a way that makes sense, either email, or if possible, uploading it back to the link
  7. Cold Hard Facts - You're sending your data to two people who will do all the work and just send it back. This isn't a fancy scheme or a high-tech operation.

Frequently Asked Questions

  • No.

  • We expect that, for pilot sized projects, this might cost a few hundred dollars. We use time as our metric for pricing. The more data you have and more complicated the annotations, the more it will cost. We can get a rough estimate for how long things will cost upon our first review, and you don't pay anything until all the data is labeled and done.

  • We annotate it... That's it. We understand that a great deal of research is sensitive, and you may not feel comfortable sharing the nitty-gritty details. That is OK! We only see what you feel comfortable sharing, so if you pre-blind your data, we won't know anything.

  • You could, however we firmly believe that data labeling is a task beneath your trainees, and adds very little to a compelling research experience. But it's not beneath us!

  • Good point! Usually, no. Our annotators are smart people, and we will do our best to do the necessary background research to understand your data. However, images which a true expert to annotate (radiology, neurology, CT, etc...) may be out of our scope.

  • We accept all common imaging formats including TIFF, PNG, JPEG, CZI, LIF, ND2, and more. We can work with any dimension - 2D, 3D, time-series, or multi-channel data. If it can be opened in ImageJ then we can work with it.

  • In theory, as much as you want, however right now we excel in smaller, niche datasets. The amount of imaging data that might reasonably lead to a single academic publication.

  • Timeline depends on dataset size and complexity. I will move hell and high water to get your data back to you as fast as possible.

  • Biological images can be large, and I don't want you to have to babysit a slow upload. Many people have access to a sharable link. If you do not, but would like to use our service, please reach out to chris@huronalia.com.

  • No.

  • No.

  • Perhaps eventually. Ever label will still be manually assessed by a human though.

Submit to get us started

It's non-binding and will be the only work you'll do

Google Drive, Dropbox, GitHub, etc.
JSON, XML, TXT, CSV, PDF, ZIP (Max 10MB)