Meta Releases Sapiens - Foundation for Human Vision Models

This video introduces Meta Sapiens which offers a comprehensive suite for human-centric vision tasks. It can be used for:Human Pose Estimation, Body Part Segmentation, Depth Estimation, 3D Reconstruction (Normal Estimation), Synthetic Data Training. Here are my thoughts
Schedule a Call:
techwithryanwo...
Website: techwithryanwo...
Socials:
Twitter: / ryanwongtech
Instagram: / techwithryanwong
Tiktok: / techwithryanwong
Facebook: / 100086062164038
If you would like a free consultation call: techwithryanwo...
#MetaSapiens

[@ediiiwang8559]

Many thanks for sharing the Colab link! Also, I’d like to ask if Sapiens can export high-quality 3D models in STL or OBJ format, like PIFUHD?

[@techwithryanwong]

it's important to note that while the model is highly effective in human-centric vision tasks such as 2D pose estimation, body-part segmentation, and depth estimation, it does not inherently support the export of 3D models in formats like STL or OBJ. These formats are more commonly used in 3D printing and model generation, which is beyond Sapiens' primary scope.

[@Hackersuncovered]

is there any guide for learning this? would love to learn how to use the pose estimation model. have been working with mediapipe for some time for a lot of my projects and would like to upgrade at some point. seems like the right time to do so.
also does image segmentation provide the coordinates of the outer outline?

[@techwithryanwong]

It provided the masks

[@greatsarmad5147]

man can you please drop the colab u made or used
i want try it too but when i tried to set it up it fails and i dont have the skill to debug it thx in advance

[@techwithryanwong]

colab.research.google.com/drive/1pQAL6A9goPSJfNHENrI_ktZ7xMeA3ZxC?usp=sharing