How do I add data?

You can add data to your Roboflow account by:

• using the Web User Interface. (Recommended)

• using our API. (For Pro users)

• ​forking a publicly available image dataset.

What data can be added?

Roboflow can ingest:

• various image formats, including JPG, PNG, and BMP.

• video formats, including MOV and MP4.

• various object detection annotation formats, including JSON, XML, CSV, and TXT.

Uploading Data with the Web User Interface

On the Roboflow Free Tier, all data is added via the Web UI. Start by logging into app.roboflow.com.

Once you have logged in, select "Create Dataset" in the upper right-hand corner. This will cause a dialog box to pop up.

You will need to provide three things: Dataset Name, Dataset Type, and Annotation Group.

Dataset Name

A way to refer to your collection of images/videos.

• If you're uploading a bunch of images of chess pieces, you might name this "Chess Data."

• The dataset name must be unique among your datasets. (For example, you cannot have two datasets both named "Chess Data.")

• Right now, we do not support editing the dataset name once you have created the dataset. If you must edit your dataset name, you can re-upload your data with the new name or contact us.

Dataset Type

This is tied to the type of computer vision problem you're solving.

• If you are attempting to put each image into one or more buckets, select Classification.

• If you are attempting to identify one or more objects in an image with bounding boxes, select Object Detection.

• If you are attempting to identify multiple objects in images with freeform polygon shapes (not bounding boxes), select Semantic Segmentation or Instance Segmentation.

  • If you do want to differentiate between different objects in the same class (e.g. identify two cats as distinct objects), select Instance Segmentation.

  • If you do not care to differentiate between different objects in the same class (e.g. all cats are labeled "cat"), select Semantic Segmentation.

  • These options are not currently supported, so request early access on Roboflow.

• If you are attempting to identify the locations of important components in an image, select Keypoint Detection.

  • This is not currently supported, so request early access on Roboflow.

Annotation Group

Annotation Group should be the broader class of objects you're detecting or the collection of categories for your classification problem. It is a way to refer to all of the objects or labels in your images.

• For example, if you are attempting to identify pawns, rooks, kings, and queen pieces on a chess board, then you might write pieces.

• If you're attempting to classify handwritten images as being 0, 1, 2, ... 9, then you might write numbers or digits.

You can learn more about annotation groups here.

How do I upload videos?

You can upload videos in the same way that you uploaded images! You will drag and drop them into Web UI.

Once you drop them into the Web UI, a dialog box will ask you to choose the frame rate. This will tell us how many images we should sample from your video per second.

• The highest frame rate you can select is 60 frames per second, which will generate 60 images for each second of video you upload.

• The lowest frame rate you can select is 1 frame per 60 seconds.

After sampling from your video, the images will appear as if you had originally uploaded images to Roboflow.

Uploading Data with the API

Uploading data with the API is only available for Roboflow Pro customers.

If you want to add data via the API, you will first need to create an API key on your account page.

Once you have generated your API key, check out our Upload API-specific documentation to get set up.

Accessing Public Datasets

Roboflow has made available over 30 (and counting!) public image datasets for you to use. From packages on porches to website screenshots, we host diverse datasets that can be helpful to you.

• If you want to get started without uploading any data, the public datasets are a good way for you to get the feel of Roboflow.

• If you are tackling a problem -- like pothole detection -- and don't have enough data on your own, you might consider merging your pothole dataset with our pothole dataset. Using public datasets is one way to improve your computer vision model.

To take a public dataset and be able to apply your own preprocessing and augmentation steps to it, select the dataset, then click "Fork Dataset" in the upper-right corner.

If you don't need to modify the dataset in any way and you don't want to use Roboflow's one-click AutoML solution, but you'd just like to export the dataset, then click the version you want on the left-hand side underneath "Downloads" and export that version in the format of your choosing.

Want to make your dataset public? Reach out to us! We'll give you access to some of our beta features or paid features.

Your data is yours!

We strongly believe that your images and videos are yours. That's why, when you upload those, your photos and your videos remain yours -- we do not own it. (You can check out additional details in our terms of service, item 22B.)