DensePose: Dense Human Pose Estimation In The Wild (CVPR 2018 Oral)

Рет қаралды 194,503

Alp Guler

6 жыл бұрын

Project page: densepose.org
Paper: arxiv.org/pdf/1802.00434.pdf

Пікірлер: 39

@LEDchip 5 жыл бұрын

This is truly wonderful work! Super kudos to the engineers making this happen, they are bringing us all into a new world, literally and figuratively. Being able to take video and capture the sets and actors in 3-d in a form that is truly useable, would allow virtually unlimited potential to redevelop story lines and re-render new movies with different actors, filming locations and may be even the way the films end, all at will. Computer farms on the cloud could give the power to anyone with a cell phone. I can’t wait to try this soon and see how it really works with my video.

@punachaa Жыл бұрын

Great Research. If not we humans..some other brains are already seeing us ( people on earth) via radio/spectrum mapping in one or other form.. It is just that we dont know ..Thank you .

@JorgeGamaliel 6 жыл бұрын

Awesome! With generative adversarial networks many things become possible.

@DamianReloaded 6 жыл бұрын

This is awesome. With this technology (and a few already existing others) they could potentially re-create any video/film recording in 3D. Imagine watching casablanca as one of the extras through VR! ^_^

@alistermunro7090 6 жыл бұрын

Casablanca is probably last on the mind with VR ;)

@alialtaf3412 6 жыл бұрын

I am in love with this technology.. This can revolutionize Motion Capture while also saving cost.

@example.com. 6 жыл бұрын

Awesome work

@hyunkim2172 6 жыл бұрын

Just wonderful!!!

@hakopato 6 жыл бұрын

Great work!

@Tactic3d 6 жыл бұрын

great work !

@757beko 6 жыл бұрын

Wonderful!

@delikatus 5 жыл бұрын

Am I correct that the output provided by the densepose code does not aim to be analogous to that of a full blown motion capture system, but instead only provides the positions of body parts in the 2D coordinates of the input images, i.e. it doesn't allow to track positions of body parts in a 3D space? Would love to be wrong about this and learn how to obtain 3D representations. Or is there other software that performs this step?

@cangulec 6 жыл бұрын

Amazing

@Oakbeast 5 жыл бұрын

how can the U / V coordinate data transfer to a bones or skeletal rig in say Maya, 3D max or Blender or any hub tool for that matter? killer tech; cant wait to see more

@khaerulumam42 6 жыл бұрын

Awesomeeee!!

@erimi01 5 жыл бұрын

Hi! I'm trying to try out DensePose myself but are having issues following the install guide on the git, is there any way you could help?

@kamillatocha 5 жыл бұрын

someone definitely needs to make a music video with this in it

@souslicer 6 жыл бұрын

please release the dataset

@SeaCreature_ 6 жыл бұрын

I love it, what a great accomplishment. Give us the GitHub link for it :)

@royaramezani7059 6 жыл бұрын

Awesome work! Can you please share the GitHub link?

@BDCCN-TranPhuongNam Жыл бұрын

hi , how can i make a densepose RCNN part index image , its image will look like second image at 1 min 4 seconds of video ?

@user-vg7zv5us5r Жыл бұрын

And then? What was that software made for?

@KD_Kartik 4 жыл бұрын

hi its awesome technology and i need help me can you tell how to install this project in system

@thegreatskywalker 6 жыл бұрын

Amazing work!!! Will the source code be available?

@jibinjohn8263 5 жыл бұрын

You can use it for research not for development. Else you have to buy the licence once facebook release the 3d pose library.

@aprendappdcomitalo953 Жыл бұрын

1984 - it´s all U need to read now

@jonhsonbuchanan1126 6 жыл бұрын

Can this be done in realtime?

@kazioo2 6 жыл бұрын

From the paper: "a discriminatively trained model can recover highly-accurate correspondence fields for complex scenes involving tens of persons with real-time speed: on a GTX 1080 GPU our system operates at 20-26 frames per second for a 240 × 320 image or 4-5 frames per second for a 800 × 1100 image."