Higher density (imported?) meshes choke new mesh mod tools

[Anonymous]

A high density mesh (10s of thousands of facets per mesh body, dozens of such mesh bodies in component) will choke the new mesh modification tools most of the time. The error is usually "<something> failed due to invalid input".

Commands that fail include: remesh, reduce, mesh to brep, and probably others. I have had limited success with "reduce" on lower poly count meshes. Additionally, the "preserve sharp edges" command is sometimes not respected. 

I am happy to share this model with A-desk staff in confidentiality as it is a licensed model import. Please let me know how best to share if interested. 

[PhilProcarioJr]

@Anonymous

All the stuff you just listed is why I told the dev team that the new mesh tools were a waste of development time and resources. Not to sound mean to the group responsible but the only thing the new mesh tools are good for is the casual maker and not good for any serious CAD work. I have tested the tools as much as is possible given the current state of Fusion and really see no benefit if your trying to achieve CAD accuracy with the tools. The tools look neat in the videos but all the stuff I have seen you might as well throw CAD accuracy out the window.....

[Anonymous]

You have to break a few eggs to make a good omelette, any omelette in fact....!

[PhilProcarioJr]

@Anonymous

You are correct but if you don't have a spatula or a frying pan there is no point in trying to make an omelette. 

Currently in Fusion there is no G2 or G3 for surfaces, T-spline tools are lacking, sketch solver is slow.... What good are mesh tools if the resulting body or surface can't be controlled and is inaccurate?

And another case in point is the current mesh space broke most of the mesh to B-rep and mesh to T-Spline workflows and as if that wasn't bad enough made performance with meshes and T-Spline terrible. Mesh space environment shouldn't have been released until it could handle millions of polygons...not thousands. Autodesk has apps like Mudbox which can handle millions of polygons in realtime..they would have been better off incorporating code from one of their apps like that so they could handle high resolution scans and models. Higher polygon counts = more accuracy. Any tools that are incapable of being accurate have no place in a CAD app. After all that's why we use CAD apps is for the accuracy.

Just my 2 cents...

[Anonymous]

as a casual maker 

and someone who has discussed with you before the use cases I am pursuing

I support your thoughts on improving the code base iteratively over time using IP from other AD products which may be more performant, 

but I dispute the notion that the ability to generate a toolpath from mesh-based bodies which is is not perfectly CAD accurate is not a valid objective.

[Anonymous]

In fact, I once made an omelette on a hot rock with a stick...just saying' ;-}

[PhilProcarioJr]

@Anonymous

"I dispute the notion that the ability to generate a toolpath from mesh-based bodies which is is not perfectly CAD accurate is not a valid objective."

Your no closer to this now then you were before...even with the new mesh tools...

The tools I was trying to get them to implement would directly give you what you were asking for. These new mesh tools do not. 

The points I was trying to make about accuracy has no bearing on what you want, on the other hand being able to convert meshes to T-Splines then to accurate surfaces does. With this ability you would not be restricted to using low poly count models for your work, so in the end you could use that million polygon mesh instead of a 10,000 polygon mesh for exactly what your trying to do.

I have been CNC machining the stuff your doing for years so trust me when I say I know exactly what tools are needed to make your work easy and smooth. I wish people could see that they are asking for the wrong things....anyways I'm beating a dead horse so I will close with this, I hope the new mesh tools help you and your work.

Cheers

[jakefowler]

Hi @Anonymous,

If you're able to share some data that exhibits the issues you're describing, we'll look into it right away. You can email to jake.fowler (at) autodesk.com.

The mesh editing tools are still a Technology Preview, meaning that we most certainly intend to add to and improve the toolset (primarily based on the feedback you guys give us!).

Many thanks,

Jake

[TrippyLighting]

I agree as usual with @PhilProcarioJr's assessments with the exception that I think the horse ain't dead quite yet. It might be on sustained life support, but it's still breathing 😉 

Agreed. In order to make an omelette you have to break eggs. However, If you make omelettes for a living and you break the pan, or the spatula - after all Fusion 360 is a tool, not a consumable! - you have a real problem on your hands.

The mesh workspace is still a technology preview. It is entirely unacceptable for a professional tool, which is what Fusion 360 aims to be, to implement a technology preview in such a fashion that reduces the functionality and performance of current productive features used and relied upon in professional environments. Even more concerning is that this happens in form of a forced update totally out of control of the user. Based on Forum posts this weekend it's taken several people and organizations out cold without warning.

Totally and entirely unacceptable!!!

The ability to generate a toolpath from mesh-based bodies which is is not perfectly CAD accurate is a perfectly valid objective. The question is, however, if

• the current mesh workspace even allows you to do that and
• whether that ability actually needs to be integrated into a CAD tool at this time when other tools are lacking.

The mesh workspace deals ONLY with triangulated meshes, which by definition lack precision. On the other hand Fusion 360 strength as a product design tool offers a combination of precision tools with two areas needing work in order to be able to compete with other available software tools. Surfacing is not nearly what it could be and for more complex T-Spline modeling one better resorts to their favorite subdivision modeling tool do do the bulk of the modeling ( or use Rhino ).

The latter I find somewhat disturbing because Autodesk owns two of the most widely used and mature tools in that area, namely Maya and 3DS Max.

[Anonymous]

Fusion is a new way of doing things. I've been developing and using 3d and other graphics software probably as long as anyone on this forum, more than 35 years, starting on HP9000, Evans & Sutherland Picture System, then Silicon Graphics IRIS machines running Motorola 68020s in the very early '80s, where actually having a hardware z-buffer, a mouse, and volumetric solid objects (shaded too!) were extreme innovations. Given that context, witnessing so many software and hardware concepts, capabilities, and architectures ascend and be regularly superseded and disrupted over these decades, I am able to contextualize Fusion360's development in a grand trajectory of graphics and 3d software workflows and UI innovations over the years. I see Fusion360 as quite obviously still advanced beta, perhaps even version 1.0, but not as bullet-proof and time-tested production v3 to v15 or whatever software. It is certainly also an unbelievably fantastic beacon of promise and well-executed innovation, bugs and all.

Anyone who relies on Fusion360 solely for the their eggs and bread should understand this platform is fresh and tasty goods, but it may not be as filling as you require for the kind of business you are in, with the kinds of clients you might have. I've been there, seen it before so many times. Autodesk is taking us to new places with its exciting new paradigm, and if you want to come along for the ride you should understand and accept the risks of surfing on the bleeding edge with such fancy new software so full of promise. However if you need steadfast industry-leading reliability, you must go with the the tried and true, not the flashy exciting new. 

It's like Tesla and their autopilot getting flack because someone finally died behind the wheel. That death was always going to happen, statistically speaking, and we still don't have the data to state the probabilities of when the next deaths will be. There will have to be more autopilot deaths to get that data. Sounds awful, but we actually do already know that even today's autopilot is very likely to cause fewer deaths per mile than business as usual (humans driving), so we still go along with it, most with great enthusiasm. Risk is part of advancing both life and business. Yet, pitifully, Consumer Reports calls on Tesla to pull its autopilot software until it can be perfected. That only shows that Consumer Reports has no idea how this complex, fantastic new software paradigm, Autopilot, MUST actually be developed if we are to advance the state of the art at all. People will HAVE to drive millions and billions of miles in AutoPilot to collect the data to make it safer and safer yet. Thats's actually the only way to do it.

Back to Fusion360: It's the same deal. There will be blood, but that's the price of progress. If you rely on it for everything, you have taken what I call a high-beta risk with your business, a risky risk but with high reward potential. Understand that. In any event you may want to always have a backup plan and not make business-breaking commitments based on this fantastic, but still fresh software. I'd say, even given the unprecedentedly high level of attention and support this software is getting from the AutoDesk team, that this software will remain somewhere around advanced beta or v1.0 for at least another year. Yet I've decided to go all in with Fusion360 because I can understand and accept the risks and rewards of being on the bleeding edge. That's me. I suspect it's many of you, too, and if so, then keep calm and carry on!

[cekuhnen]

@Anonymous

Obviously Fusion 360 is new - but it is not that new anymore. It is one the market since quite some years now.

We all can also understand and agree on that Fusion 360 tries to offer more than just a modeling environment such as physics cam online collaboration and others.

To get all those modules right it will take time obviously.

And that is the point I want to focus on in my argument. The mesh tools are a tech preview. So I am ok with the fact that currently it is not as powerful or polished.

However since 3 years do I suggest to AD that Fusion 360 has insufficient modeling tools.

Surfacing:

Loft needs to have rails to be able to blend between two surface edges well

Loft offers no G1 G2 along edge rails

There is no square tool - the patch tool creates a terrible draped surface when doming if you want a cleaner surface

Patch offers G0 G1 G2 but does not give you the ability to adjust the G1 G2 weight nor specify the edges with individual G1 G2 settings.

Loft solid and loft patch can have different results

Fillet offers G2 but has no G2 weight slider

Loft G1 and G2 however offer this

Extrude surface only works alone perpendicular to sketch plane

and others

Timeline:

Move command is not really parametric because the feature does not save and expose the values you used when creating the feature

Align command results only in to a move command.

Inspection:

There is no ability to project curvature combs over a surface along U and V or a specified direction and angle

There is still the problem that with a too big scale a G2 comb breaks making you think that G2 is not valid

Sketch:

We still do not have CV curves to create better profiles and easier create blend curves

Spline G2 constraint only gives you access to the G1 tangent handle and you cannot massage the curve with modifying the G1 and G2 point

You have add dimension handles onto the G1 tangent handle and make tiny steps to and observe the curvature graph resulting into a massive amount of value entering and checking cycles

We still only deal with 2D sketches and the 3D sketch approach is not very usable.

As a serious industrial design Fusion is still ill equipped to tackle many surfacing needs which is why in my teaching we still use Alias to do the outside surface

and then use Fusion to do the inside design. This is also not an unusual approach and also gives me the ability to educate my students more about CAD workflows than just learning a

parametric package like SW or Fusion where you have to work in a specific way. Apple uses Alias and NX and other companies use that workflow as well.

You also model concept models and manufacturing models you might also want to use different tools anyway.

Currently I work on two medical projects and actually use Blender for the ideation phase before I go to Alias and then Fusion.

But Fusion got advertised as "Design differently".

And that it is just plain annoying when certain things work genius and others literally break it.

Fusion 360 has a fantastic solid modeling toolset. 

The rib and web, draft adjustment, boolean tools etc are just great and the timeline is genius.

Certain surface structures Fusion also solves so much more elegant and easier than Alias which forces you to do proper surfaces in a Class-A matter which is for technical surfacing

needed and good but honestly for many designers just too much in terms of how hard it is and how labor intensive it is.

Specifically rounding tips is a tough thing.

This is one of those examples where Fusions 360 loft tool is just fantastic with what it can do:

But without G1 or G2 along the rail pointless.

I am quite sure that the devs who worked on the Mesh module are not the same that work on T-Splines are not the same that work on surfacing.

However and I think the criticism here is right where does AD put the focus onto.

I rather would have less diversity but the current tools be made perfect before doing to something else.

In the end I think it would also help Fusions reputation because I got the notion from few in the industry here in Detroit that they think Fusion is a cool

lets play with an idea CAD tool but for serious work they still need and have to rely on apps like SW because they have the tools Fusion just does not.

[SimonPlatten]

I've downloaded Fusion 360 from the apple app store version 2.0.3800.  I have a model with a Facet Count of 160771, I've tried to reduce the mesh several times by selecting all, then in the reduce dialog:

Mesh Faces / Body:  1 selected (this is the entire mesh)

Reduce Type:  Adaptive

Reduce Target: Density

Density: 0.2

When I click on Ok, the result is the same every time: Reduce failed due to invalid input

What can I do to resolve this?

[TrippyLighting]

Use Autodesk Meshmixer for working with meshes.

[SimonPlatten]

Thank you, I will take a look now.