V7 - Image Dataset Management

Load any image or video format

Supporting JPG, PNG, TIF, MP4, MOV, SVS, DICOM and counting. Upload data for video understanding to ultra-high resolution histology slides and browse them from the same platform.

Track annotation progress

Image status dynamically changes as your project advances, allowing you to train models on completed sections at any time.

Rainforest airplane n2.png

New

Assigned

In Progress

Pending Review

In review

Complete

Amazon_rainforest_aerial.png

Complete

Amazon_rainfor.png

Complete

Search for the right image

Filter by upload date, annotator, tag, classes, and labeling progress. Preview images in a flash, powered by a high speed Elixir back-end.

Balance your data

Take a glance at your dataset's composition. Troubleshoot your model's performance by taking a look at class representation and instance frequency.

Collaborate

All the data scientists you work with should know exactly what your ground truth looks like. Invite unlimited team members, annotators, and collaborate in real time.

Version control

Never lose track of your annotation versions as your dataset evolves. Reference each model to a dataset version.

3.0

20/05/2020

317k Images

14 Classes

2.0

13/04/2020

317k Images

12 Classes

1.1

28/3/2019

224k Images

6 Classes

1.0

12/01/2020

172k Images

6 Classes

Developer Kit Included

$ darwin create example-dataset
Dataset 'example-project' has been created.
Access at https://darwin.v7labs.com/datasets/example-project

$ darwin upload example-dataset -r path/to/images
Uploading: 100%|########################################################| 100000/100000 [00:01<00:00, 2.29it/s]

$ darwin pull example-project
Pulling project example-project:latest
Downloading: 100%|########################################################| 100000/100000 [00:03<00:00, 4.11it/s]

from darwin.client import Client

client = Client.local()
client.create_dataset(name="My New Dataset")

from darwin.client import Client

client = Client.default()
dataset = client.get_remote_dataset(slug="example-dataset")
progress = dataset.upload_files(["test.png", "test.mp4"])
for _ in progress():
print("file uploaded")

from darwin.client import Client

client = Client.default()
dataset = client.get_remote_dataset(slug="example-dataset")
progress, _count = dataset.pull()
for _ in progress():
print("file synced")

Load images and annotations directly into your favourite framework

from torch.utils.data import DataLoader
from darwin.torch.dataset import Dataset
‍
# Instantiate your Darwin dataset
dataset = Dataset('team-name/dataset-name', validation=0.2, test=0.1)
# Use it in a simple PyTorch dataloader
dataloader = DataLoader(dataset, batch_size=16, shuffle=True)