Can someone use this on Elon's "my heart goes out to you" video, and compare it to... similar gestures. Would be interested to see exactly how close it is.
Looks cool. How’s it different from https://github.com/gaomingqi/sam-body4d ? My guess: Is sam-body4d focused on temporally consistent body reconstruction into an MHR model, and Lift4D creates the full scene and a model of the tracked object (that doesn’t need to be a human). I may be way off, but a comparison would help me!
wondering how accurate the extrapolated distances are? Like the Rhino and the tree -- I wonder if the approximation could be useful in something like forensics from a security video?
> I wonder if the approximation could be useful in something like forensics from a security video?
No. It's just some plausible thing the AI made up. It's literally like asking an LLM to generate evidence of a crime based on the prompt you presented.
There are plenty of cool use cases, just not forensics.
Used properly I could imagine legitimate forensics uses to 3d reconstruction and splatting. For example as a "map" to find what underlying data (e.g. video frames) captured the thing.
Cool, so now when the swarm drones come after you, they can reconstruct the video on the fly from that single CCTV shot taken by flock, and pinpoint your location from the aggregated data by plantir!
https://youtu.be/TV0ZUv4CMJ8?si=53etD-IUhsQkPpHt&t=1m41s
No. It's just some plausible thing the AI made up. It's literally like asking an LLM to generate evidence of a crime based on the prompt you presented.
There are plenty of cool use cases, just not forensics.
It's only a matter of time before it's in use by law enforcement: https://news.ycombinator.com/item?id=48520807