philipp retweeted
Excited to introduce HypLL, our brand-new hyperbolic learning library developed together with @phippli and @PascalMettes!
1/4
philipp retweeted
Two years ago we released our open source framework for self supervised learning github.com/lightly-ai/lightl….
Today it's used by leading tech companies and research institutions all over the world, reached almost 2k GitHub stars and has been installed over 100k times.
philipp retweeted
We added support for the Masked Siamese Network (MSN) to @lightly. MSN learns image representations by comparing cluster assignments of masked and unmasked image views.
Great work by @mcaron31 @p_bojanowski @imisra_ @armandjoulin
docs.lightly.ai/examples/msn…
github.com/lightly-ai/lightl…
philipp retweeted
Finally published @MSSEonline
👇
We developed and evaluated a deep learning approach to automatically segment anatomical cross-sectional area in panoramic ultrasound images of lower limb muscles.
🙏 @ritpau @MVFranchi @FabioSarto3 @NeilJCronin84 @phippli
journals.lww.com/acsm-msse/a…
philipp retweeted
A friend just shared this with me. A great example for how deep learning models struggle with out-of-distribution data.
I wonder how many carriages the model has seen during training 😅
philipp retweeted
Open access code & GUI ✅ +
Instructional video 📹:
youtube.com/watch?v=It9CqVSN…
👌🏼⤵️
This Post is from an account that no longer exists. Learn more
philipp retweeted
Open source, @PyTorch, @PyTorchLightnin, self-supervision, and models from @facebookai and other labs, these are a few of my favorite things... Good luck Lightly! github.com/lightly-ai/lightl…
philipp retweeted
Collaboration with @Arm: Embedded covid mask detection on an Arm Cortex-M7 processor using @PyTorch. From self-supervised pre-training to transfer learning, conversion to TensorFlow Lite, quantization, and running it on the device. All summarized here:
lightly.ai/post/embedded-cov…
Thanks so much for sharing my blog post!
"Which Optimizer should I use for my Machine Learning Project?" whattolabel.com/post/which-o…
That's a neat quick reference table (personally, I am lazy and usually just stick to ADAM with default settings, works pretty well out of the box most of the time):
