From 2D to 3D: 4 Ways to Make a 3D Reconstruction from Imagery

In this episode of Computer Vision Decoded, we are going to dive into 4 different ways to 3D reconstruct a scene with images. Our cohost Jared Heinly, a PhD in the computer science specializing in 3D reconstruction from images, will dive into the 4 distinct strategies and discuss the pros and cons of each.
Links to content shared in this episode:
Live SLAM to measure a stockpile with SR Measure: srmeasure.com/professional
Jared's notes on the iPhone LiDAR and SLAM: / everypoint-gets-hands-...
00:00 Intro
01:30 3D Reconstruction from Video
13:48 3D Reconstruction from Images
28:05 3D Reconstruction from Stereo Pairs
38:43 3D Reconstruction from SLAM
Follow Jared Heinly
Twitter: / jaredheinly
LinkedIn / jheinly
Follow Jonathan Stephens
Twitter: / jonstephens85
LinkedIn: / jonathanstephens
This episode is brought to you by EveryPoint. Learn more about how EveryPoint is building an infinitely scalable data collection and processing platform for the next generation of spatial computing applications and services: www.everypoint.io

[@iversonvarun]

ARKit does do loop closure. I haven't experimented with loop closures in large scenes such as whole neighbourhoods. But within ( as a rough estimate) the bounds of a large room in your house, loop closure definitely takes place. To test it out, you place an AR object in the scene, walk around away from the object and then return to it, you can notice the drift being corrected through loop closure.

[@EveryPoint]

Interesting. When did you last test it? We found that that it used to early on. Then it stopped. Perhaps we need to retest on our end. We also implement our own loop closer within software we develop.

[@MrTilePRO]

Great video! What are the programs to use for this Photo to 3d video or model?

[@jonathan-._.-]

🤔 how does a slam algorithm bahave as the dataset gets bigger ? i assume one youve found enough "matches" to locate a certain image in 3d space you can abort most of the other checks , but if you add an image that isnt connected to the existing graph its gonna get interesting
would be interested in ram/cpu requirement estimations to see if i can run it on my local machine
(im trying to do drone exploration/mapping)

[@Instant_Nerf]

I’m curious what was used to 3D reconstruct from the actors from Terminator dark force ?

[@faizshaikh476]

great content 👍. I am trying to reconstruct a 3D model of a Ring and a necklace and would like to have a detailed model , can you suggest me the best way forward?

[@EveryPoint]

That is extremely tough. The materials are shiny and thin. Most likely you would need a handheld scanner typically used for modeling parts and reverse engineering.

[@AshishRaj-oc2oc]

Hey EveryPoint, I have a project on the Reconstruction of Broken Statues using AUTOCAD and Python, but I am unable to figure it out. My project deadline is at the end of next week. Could you please make a video on how to do it as it will be very helpful for my future endeavors. Thanks, mate!

[@user-bn9mj9mq4k]

have you made it? If so, how?

[@alienkishorekumar]

The video and audio are not matched.

[@mrosskne]

finally i can make my furry waifus real