Dataset Colab is a software that allows all FRC teams to submit their annotated images of the FRC game objects and other robots to a large dataset which they then can download to train object detection models. By signing up you get access to the current 2023 (and 2024 coming soon!) datasets and the ability to upload your datasets to grow the common datasets.
Benefits of Large Datasets - When training YOLO or other custom neural network based object detection models, larger datasets with more edge cases lead to higher accuracy. Currently, the 2023 dataset is the combination of all the public datasets on Roboflow, and has over 30,000 images which could never be collected by one team.
Upload Methods - Dataset Colab currently accepts any uploads in the COCO dataset format and any Roboflow datasets (simply provide the URL and Dataset Colab will fetch the dataset).
Community Support - Dataset Colab is a fully open source project (available on Github) developed and maintained by an active and welcoming Discord community. Feel free to ask any questions there!
Comparison with Roboflow
Roboflow is a great tool for annotating images and exporting them to a variety of dataset formats. While Roboflow does allow users to make their datasets open source through Roboflow Universe, it unfortunately does not allow users to collectively collaborate on public datasets. Dataset Colab fills this gap by allowing any user to contribute to collective datasets.
Current Features
Uploading both COCO datasets and any public Roboflow dataset.
Viewing all of your uploaded datasets and the combined communal dataset.
Download the combined dataset via the website or API.
Planned improvements
Upload and download datasets in the YOLO format. (January 15th)
Cool idea. Would you want to just host it on Roboflow Universe?
I’m one of the founders and we’d be happy to give you increased limits so you can combine/curate the open source datasets into a better one & collaborate with others.
Just send me an email: brad@ domain of the company
We are working to resolve this now. Sorry for the inconvenience. While the robot dataset came with the note class, the note class is never used in any of the annotations. We will remove it from the robot dataset.
Dataset Colab now has pretrained object detection models! Download the model weights and follow our documentation to start detecting notes and robots. We currently have 4 pretrained models:
YOLOv5n (95.92% mAP)
YOLOv5s (96.01% mAP)
YOLOv8n (96.25% mAP)
YOLOv8s (95.60% mAP)
On 2/1/2024, we will be releasing Limelight + Coral compatible (tensorflow) models: ssd-mobilenet-v2, efficientdet-d0, and ssd-mobilenet-v2-fpnlite-320. Signup to get an email notification when these are released.
Update: Dataset Colab now has only note and only robot models in addition to the note and robot models. Visit the models page to download and run these new models.
Dataset Colab now has Tensorflow object detection models (SSD Mobilenet v2 & EfficientDet) that can be used with the Google Coral and a variety of other accelerators! To download and get started using these models visit our models page to download the model and our documentation page for a guide to get started.
SSD Mobilenet v2 (92.20% mAP50)
EfficientDet (94.50% mAP50)
New Only-Note and Only-Robot Models
In addition to these models we also trained only-note and only-robot models. These models are more specialized and have greater performance. To download and use these models visit our models page .
Recommend New Models
If there are any other models that you would like to see on Dataset Colab, DM me and we would be happy to train them for you!
