Data labeling is time-consuming and can be very expensive. Lightly tells companies which data to label to have the biggest impact on model accuracy while saving time and costs.
Hi Product Hunt! Weβre @matthias_heller and Igor from Lightly.
Lightly is a platform for companies training machine learning models. Most companies that do machine learning only label about 1% of their data. We tell them which 1% of the data to label to have the biggest impact on model accuracy.
One of the biggest challenges in every machine learning project is curating and labeling the collected data before training the machine learning model. We want to democratize machine learning by building a platform to boost engineers to overcome this hurdle. Lightly helps to maximize model accuracy by finding the right data to label. As a data curation platform, Lightly offers tools to reveal and eliminate common problems like data biases, imbalances, and redundancies.
To help you work with unlabeled data we leverage recent advances in self-supervised learning. We are also maintainers of the popular open-source framework lightly (https://github.com/lightly-ai/li...). We work with companies ranging from small computer vision startups to large companies working on robotics and autonomous driving.
Weβd love to get feedback from you and will stay around to answer questions all day!
Our website: https://lightly.ai/
Our application: https://app.lightly.ai/
@matthias_heller@igor_susmelj so great to see you guys launch and catching up the public on a little bit of the magic you are building. Canβt wait for them to see the rest π
Report
This is definitely a must for data scientists looking to increase the performances of their models by curating their data! Well done!
Nice!!! Data curation is such an important but way too underestimated part of developing a good ML solution - partially because it is so time consuming! This is a great solution to save time, improve the quality of ML models and reduce bias in the data. Great job!
@andrew_mertell Glad to hear this! If you need any help in trying it out or have questions just let us know. We are happy to assist :-)
Report
All I hear are good things internally about the Lightly Team. They are very helpful and experts in this domain.
My team today saves hours every day with lightly. We are all excited to continue working with you all!
I highly recommend them anyone working on training models with large datasets. πππ
Lightly
Lightly
Lightly
Veezoo
Lightly
Lightly
Lightly
Dashblock
Lightly