One thing I started a few years is multicam walks. This means that I carry some 4-6 Osmo Pockets to actually film in several directions simultaneaously - example: Dresden [2].
The editing process is taking multiple hours so I have hundreds of unreleased recordings. If someone has a hint on AI assisted multicam editing, where to start (like better gaze prediction etc. from VR) I'd be interested - especially feeding all camera streams and let a NN decide the best camera angle and cuts to be between 4-11s?
Maybe it's time to start an open source effort to do something similar in some corner of the world? You'd need access to the mapping camera hardware, obviously. Who would be willing to fund it?
I guess subtitles, HDR and decent editing are more important for now so I'm looking for AI help in the editing part. Subtitles are already supported by GPTs
Thanks for making them, will definitely watch some!
It's legal to take pictures in public places in the Czech Republic, no need to ask permission for that.
This dataset is the result of many, many hours of actual driving, with our own camera hardware (mosaic X, designed and built right here in the czech republic) and dual RTK + GNSS/INS/VSLAM system to allow extremely precise positioning.
I would like to encourage adhering to the CC-BY-NC-SA license as much as possible. I would like to at least have a quick dialog which each person who grabs this dataset to explain this as clearly as possible.
This data is not public domain. It was created through tremendous effort and sunk cost, and I would like people to use it for research / learning purposes and to share their results with the world. Releasing all the data with a public link is honestly a bit too scary for me at this point.
Hence the signup form :-)\
After you sign up, you'll get a link to the dataset. First a link to a very small dataset, which is all we have managed to anonymize so far (we have to obscure faces and license plates in order to adhere to GDPR legislation). We will make the entire dataset available in a few weeks.
You might redirect your critique to the other ~1000 companies in the world who have such data and don't release anything at all under any terms.
So far we have had a few hundred signups for the data. I ran the applications through chatGPT to get an idea of what people want to do with the data:
Research and Academic Purposes: A significant number of respondents plan to use the dataset for research in areas like computer vision, 3D reconstruction, Gaussian splatting, neural rendering fields (NeRF), and autonomous driving. Many aim to publish papers, conduct experiments, or work on academic projects such as theses.
Learning and Educational Use: Many individuals expressed interest in using the dataset for learning purposes, personal education, or as a teaching aid. This includes students working on class projects, hobbyists looking to enhance their skills, and educators seeking material for
lectures and demonstrations.
Personal Projects and Hobbyist Exploration: A considerable portion of respondents wish to explore the dataset for personal enjoyment or hobby projects. This includes experimenting with new technologies, exploring the city virtually, or simply satisfying personal curiosity.
Virtual Reality (VR) and Augmented Reality (AR) Applications: Numerous users intend to integrate the dataset into VR/AR environments using platforms like Unreal Engine and Unity. Applications range from creating immersive experiences and simulations to developing VR walkthroughs and gaming environments.
Video Creation, Animation, and Visual Effects (VFX): Several respondents plan to use the dataset for filmmaking, creating animations, or experimenting with visual effects. This includes producing content for personal portfolios, social media, and non-commercial projects.
Experimentation and Testing: A number of users are interested in testing new technologies, workflows, or software capabilities using the dataset. This includes experimenting with rendering techniques, 3D modeling, and spatial data handling.
Gaussian Splatting and Advanced Rendering Techniques: Many respondents specifically mentioned Gaussian splatting as an area of interest. They plan to use the dataset to test, implement, or improve techniques related to point-based rendering and large-scale scene visualization.
Game Development: Some users aim to incorporate the dataset into game development projects. This includes creating realistic environments, developing game mods, or using the data as reference material for designing game levels.
Artistic and Creative Projects: A subset of respondents expressed intentions to use the dataset for artistic endeavors, such as digital art installations, motion graphics, and experimental visualizations.
Virtual Tourism and City Exploration: A few individuals plan to use the dataset to virtually explore the city, create virtual walking tours, or develop tools that allow others to experience Prague remotely.
Key Takeaways: High Interest in Research and Education: The dataset is highly valued for its potential to contribute to academic research and educational purposes.
Strong Enthusiasm for VR/AR and Advanced Visualization: There's a notable enthusiasm for using the dataset in VR/AR applications and exploring advanced rendering techniques like Gaussian splatting.
Diverse Range of Uses: Respondents show a wide range of interests, from personal exploration and hobby projects to professional development and creative arts.
In the list of world cities by population, it ranks 403rd.
https://infocartography.com/world-top1000-pop
For Czechs or Central Europeans in general, it is a major city, but not in worldwide context.
I'd definitely say it's a major EU city. Maybe even a major The West city. I agree that it's hard to compare it to places in Asia, though.
What specifically would you like to use for wikimedia commons? You don't need all 1.25 million panoramas for that, do you?
We would like to know who wants the data and why. Because that's interesting, and we hope to make new relationships out of this. Throwing the data into the void is a bit too scary as the first step. Maybe next time? :)
thanks! Jeffrey and the Mosaic team
I'm wondering if I can create them myself or should donate the many TB of multicam recordings.
This way it can help to understand for which uses Mosaic X could be potentially useful (and this could bring creativity, e.g. to develop a game).
Satellite imagery providers often give a couple of sample pictures in the public as well.