Thanks for the updates and detailed response. I am sorry to comment without having more experience with your product and awareness of it's features and history. However, as an outsider it still seems rather strange that the display system is layers of 3d arrays in which it seems a little difficult to understand distance and association. When I see this interface I think of Cooliris (http://www.cooliris.com) which lets the user view large sets of images by sliding past them like a wall. I think this could be more appropriate as the overlap would no longer be there making the visual language of the interface a little simpler. At the same time, another interesting user interface for searching data that is checked primarily by visual qualities can be found in Microsoft Lab's SeaDragon and Photosynth projects. There is a view in which all the images of one project are laid out like a wall of sticky notes and as you select one others will rearrange around based on various heuristics. With such a user experience it might be possible to create something that behaves like a genetic choosing algorithm while maintaining a permanent visual interface.

The size issues you both mentioned are obviously quite important. I imagine if need be you could do some interesting user tests to find how frequently their selection process is interrupted by a lack of perception of size.

The web is an interesting frontier. I guess there are a few large repositories of online 3d data but I can’t imagine public APIs for these are available. I suppose your service would be great as a licensable search interface for groups like 3D Warehouse and TurboSquid.

I would like to try it but I am now only working on OSX almost exclusively and I am no longer working with large sets of engineering data. Looks like a great product.

Thanks,
Mark Whiting

All good points you make. I think the simple answer is we've tried these things and opted for a 'simple' solution for now. But we're thinking that maybe in a future release we'd have a 'power-user mode' where there's the option to customise layouts and colours a bit more.

Some specific replies:

a. Searching by size. You can specify by what kind of similarity you'd like to search. 'By Similar Shape ' is the default, but there's also “By Similar Size and Shape”, “By Similar Cross-section” and a number of others.

b. Scaling of display. We played with this a lot. The problem is, as Mr Idea said below, when you've got a wide range of sizes on the one display. You end up with some parts visible and others tiny. We tried having a false scale where small parts are smaller but not as small as they would be to scale and large ones vice-versa. However everyone then misjudges sizes because they assume things are to scale and inevitably the smaller parts are harder to see.

We've tried putting rulers and other indications of length on, but they tend to clutter the display. But I agree that this is important and we are looking for a good way to do it. For now you have to zoom in on the part and then there is a ruler indicating the length.

c. Different axes for different attributes. This was the way we laid out early protoypes. The problem is that the distribution of parts in 'attribute space' is typically very uneven so you either have lots of spaces (and wasted real estate for displaying parts) or the parts have to rearranged in a way that feels wrong. Again, we do think there's merit in doing something with this, but it needs to be 'right'.

In this version, the way we choose which parts to display and how we lay them out is, perhaps, more subtle than you can see without trying it for yourself (you can try it from http://www.partbrowser.com/) . Not only the most similar parts are displayed but also some 'suggestions' which makes exploring and browsing through your parts possible.

d. Web. At the moment searches are all on either local or network drives. We may have something on the way for the web soon…

This is a first version. We know there's plenty of scope for improvements – I'm particularly keen to improve the way metadata, text and shape can be used together to search and filter so that parts can be found.

By the way, we're nothing to do with Siemens other than we have a version for Solid Edge.

Andy

This software is fairly new and my guess is Siemens will add features in the future. I'm sure Andrew Sherlock would like to hear your feedback. He's at http://www.shapespace.com/

I also think the interface could be cleaned up a little, layered 3d result sets with no visual clarification or selection preference seems strange. The layout could be made to represent specific trends in the quality of a specific search result. For instance, the horizontal axis could represent parts of similar topology, while the vertical access could denote deviation in scale.

I wonder what the semantic web version of this kind of product is. I also wonder at what point this kind of tool will be available for public 3d data, like google images for cad.