6 ms·
Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items
- achow 5y agoThe library can be seen here: https://app.ignitionrobotics.org/GoogleResearch/fuel/collections/Google%20Scanned%20Objects https://app.ignitionrobotics.org/GoogleResearch/fuel/collect...
- HNHatesUsers 5y ago
- whywhywhywhy 5y agoThe download buttons are pretty glitched, only had a few actually work.
- pj_mukh 5y agoAwesome to see the default format for the dataset be Gazebo, but doesn’t the material rendering system not compare to photoreal results from Blender? Interesting choice.
- blenderdt 5y agoIt has nothing to do with photorealistic rendering. The pictures of the scanned objects are put on the 3D model.
- dotancohen 5y ago> Interesting choice. Things like this are often not "a choice", as in twenty engineers did not sit in a room and discuss the merits and drawbacks of each format. Rather, they are "a doing", as in the intern tasked with scanning the object used the first tool that appeared in his DDG search.
- bawolff 5y agoThat's still a choice.
- dotancohen 5y agoThe point is that there was likely very little consideration involved. Thus, there is no need to ponder why one format was chosen over another.
- yreg 5y agoYou are just making that up though…
- teddyh 5y agoUnless anyone was there to witness it, all anyone can do is speculate, i.e. make things up.
- dotancohen 5y agoDid you ignore the "are often" line in my original comment? I'm talking about the process in general.
- weird-eye-issue 5y ago> that appeared in his DDG search The most unreasonable thing in your comment and that's saying a lot
- SotCodeLaureate 5y agoThese are just standard OBJ files with diffuse textures only and trivial material settings (all white). Additional there is some metadata that specifies that the same data is to be used as displayable model and collision geometry. So, essentially, this data is highly portable, photoreal aesthetics is not really a goal. (Also see some notes in the paper, e.g. "Limitations of the Dataset".)
- pj_mukh 5y agoWell now I'm curious, is the claim that photoreal-ness doesn't matter when training network for the real world. Weird.
- SotCodeLaureate 5y agoRemoving lighting before object recognition may be a separate step, yes. But in this case it's just that their goal is to scan objects, shapes as such, and not intrinsic properties of their constituent materials (such as specularity).
- radarsat1 5y agoI think it's great to see Gazebo acknowledged here. I've often wondered why it's not more often considered for the basis of RL systems, as it is designed for robotic simulation, and supports multiple physics engines.
- peteforde 5y agoI got excited, but now I am just confused. How is this novel? There are far higher quality object scans available on sites like Sketchfab and Turbosquid, not to mention free assets on the Unity and Unreal asset stores. And possibly even more perplexing: why is it presented as an academic paper?
- spoonjim 5y agoIt doesn’t claim to be a breakthrough - it’s just a set of assets where Google has set a minimum quality bar, you’d have to evaluate stuff on Sketchfab for quality if you wanted to use it.
- byecomputer 5y agoThe level of detail seems variable. For example, this one looks like a Counter Strike 1.6 map: https://app.ignitionrobotics.org/OpenRobotics/fuel/models/Fortress%20island https://app.ignitionrobotics.org/OpenRobotics/fuel/models/Fo...
- jorams 5y agoThat model doesn't seem to have anything to do with the Google Scanned Objects dataset.
- fxtentacle 5y agoI guess they want people to be able to cite it, which is why they wrote a paper. Also, many people still use ShapeNet (for example for pre-training optical flow algorithms before the self-supervised stage) and this is an obvious quality improvement over that, albeit with much fewer models.
- legel 5y agoThe thing to know about 3D scanning -- and more generally, automated photorealistic 3D modeling of the entire physical world -- is that it is an unsolved problem. A lot of the 3D scans you'll find were created with techniques (e.g. photogrammetry) that do not scale to a majority of the physical world. The concern with assembling datasets for ML from e.g. Sketchfab and Turbosquid, besides typical copyright restrictions, is very high variability in 3D model quality. I love that Epic Games indeed seems to be very excited about 3D scanning (acquiring several companies), and their Unreal Engine does offer a large number of free 3D scans. But the scope of their publication so far is just a drop in the ocean of possibility for 3D computer vision. Even if I may disagree with and see further opportunities in the techniques employed for 3D scanning, I commend the Google authors on their hard work over multiple years collecting massive datasets, their detailed academic communication, and their open source licensing. In my view, anyone working on and communicating new techniques for 3D scanning/modeling/vision like this, and especially anyone who is publishing large numbers of 3D models for the world, is contributing to a much bigger future, where we can have the entire world fully 3D modeled.
- changoplatanero 5y agoI remember seeing a demo of this when I was a google intern in 2015
- HNHatesUsers 5y ago
- rbanffy 5y agoI wish they did that with the Computer History Museum's collections. Or the Living Computer Museum in Seattle. Maybe even paying them to help them continuing to exist. Extra bonus for doing this with the disassembled parts during restoration so everyone can 3D print their own Cray 2 aquariums.
- rbanffy 5y agoWhere do I report a bug? https://app.ignitionrobotics.org/GoogleResearch/fuel/models/Threshold_Porcelain_Teapot_White https://app.ignitionrobotics.org/GoogleResearch/fuel/models/... is not the Teapotahedron.