Saturday, February 3, 2018

Colgate EALL

Colgate University, located in upstate New York, with nearly 2900 students, ranks in the top 15 in the country. A while back the department head of the East Asian Language Laboratory contacted me with a project. The wanted to remodel their department, consisting of 5 faculty offices along a hallway, with a small room at one end and an even smaller niche at the other, to better reflect the ethos of the department. They wanted one room to be modeled so as to give a Chinese feel, and the smaller niche to have a Japanese motif, and for both areas to provide a place to people to sit and relax while waiting to meet with professors.

The provided me with some measured sketches of the various rooms, photos of the sorts of details they were looking for, and some of the building blueprints. Of course, building blueprints and ‘what actually was built’ can sometimes vary, and trusting the room dimensions taken and jotted down by someone who does not normally do such things can be unwise for a fabricator. Best to confirm the numbers for yourself before getting too far along. But given the irregular shape of the rooms involved, it was all too possible as well to go to site and make an error in taking dimensions somehow, either writing down the wrong number, or ascribing a dimension in one place to another, or just forgetting to measure something that later on proves to be important.

Now, Colgate is some 4 hours drive away from where I live, so the prospect of making several 8+hour trips back and forth to take dimensions was, well, seeming inefficient at best. I had recently however received some information from Trimble (the company which owns SketchUp) of a new 3D scanning product called Canvas IO. Here’s the video:

Sure looks pretty whizz-bang, huh? It certainly seemed to offer advantages over physically going to site and manually taking measurements.

Looking over the Canvas IO website, they make the following claim for accuracy:

Across the case studies we’ve run with professional out in the field — using Canvas on real-world projects — we see that most measurements are within 1-2% of what’s verified manually by tape measure, laser range-finder, or existing blueprint. 

So, after looking it over and seeing the possibilities, and liking that direct export from the captured data into SketchUp was offered, I thought it would be a good way to go. Unfortunately I do not have an Ipad, but a few emails later I had managed to convince the Colgate IT department to purchase the device ($500) and to go and scan the room.

It took a few weeks, but eventually I received the first scan, which opened in SketchUp like this:

It looked good! The depiction looked faithful to the room layout I had seen in other drawings they had provided.

The ceiling of the room was saved as a group or component, so one click to disappear that part and the contents were revealed:

That also looked pretty good.

Now, at the preliminary phase of the design discussion with Colgate, I had been suggesting they might want to look at using Japanese and Chinese ceiling treatments as a way of defining the spaces, so I was curious to see what details, like the locations of sprinklers, HVAC outlets, special access hatches, and so forth, might be shown on the Canvas Scan as that would affect what sort of ceiling might be the best choice. Unfortunately, the ceiling component of the scan was completely featureless. This is a typical metal framed ceiling with acoustical tiles, so it is not completely featureless, and I know from other supplied drawings that there are HVAC ducts in several areas.

Thinking that the people at Colgate had omitted to scan the ceiling, I asked them to do another round of scanning, and a few days later I received another CAD file. It was no different however in respect to the ceiling - a blank sheet. I asked the IT person at Colgate if he could talk to the people at Canvas and ask them why this info was missing from the scan. I mean, either it scanned or did not, which would show in the raw data from the scan, and if it did show in the raw data, then why were they choosing to not detail that part in the converted file?

Canvas’s site explains that:
Your CAD file will be an architectural “shell” of the space captured. Doors, walls, ceilings, floors, countertops, cabinets, and other “built-in” features will all be included, but no movable objects like furniture and decor. 
Emphasis mine.

That statement turns out to be misleading, at least in terms of ceilings. And when Colgate’s IT person tried contacting the company with questions about the scan, all they did was refer him to their FAQ page.

It’s like a lot of tech companies these days. They love to sell you a product but if you require any help afterwards, they either outsource the tech support to India, etc., or they just provide a FAQ page. Forget actually providing customer service, someone with whom you can have a conversation. I guess that costs too much. Grr..

Anyway, eventually the idea to employ ceilings was dropped due to budget constraints, so that aspect became less concerning. In the past several days, having received a drawing deposit from the University, I’ve been busy sketching.

Everything has been rolling along fairly smoothly, however yesterday I noticed some curious things.

First here is the room at one end, which will be the Chinese-themed room. This is a view down the hallway to that room:

Notice the window with the heater below? Note their relative sizes to one another - the window being narrower.

This however is how the window and heater actually look:

Clearly the heater is in fact narrower than the window. Why is the electrical box next to the heater missing from the scan? Notice also that the top of the window opening is a couple of inches below the ceiling, and a roll-up blind is placed in the junction. The scan however shows the window going right up to the ceiling.

Worse yet, the scanned window dimensions do not match the directly-obtained dimensions I had from Colgate. The window actually measures 53.875" wide, but the scan has the window at 52.36". That’s not a tiny difference and would be outside the range of error one might anticipate from a person taking measurements with a tape measure. Is it within the company’s stated “1~2%” range of error? no, it isn’t, the difference between the numbers is close to 5% off.

Then there is this curious anomaly on the scan, as we look from the window ledge back into the room itself, we can see that the window trim piece sticks up in the air:

At first I thought I must have accidentally grabbed it with the mouse and pulled it up, however re-checking the original file they sent, as well as the second scanned and converted file, revealed that both showed the window trim the same way. Not sure what is going on there. Notice also in the background how it shows the baseboard. It measures, in the sketch, as 0.5" thick and 4" tall, however the actual baseboard is not wood but glued-on vinyl, which is about half that thickness at best.

It gets worse. The sketches I had from Colgate indicated that the hallway was 59" wide, however the scan from Canvas showed a hallway that was 55.1186" wide. Such an accurate number, but very different than what was indicated on Colgate’s materials. Again, I checked both scans to see if one of them showed something different, but no. Then I contacted the Architectural Trades Manager at Colgate and had him go and double check the measure for the width of the hallway, which again came back at 59". So the scan was giving a number which was nearly 7% out from actual. That’s terrible!

I am now completely disillusioned with this product. It doesn’t perform as advertised and cannot be relied upon. I feel bad that Colgate purchased the device upon my request, given the results. While I will be able to navigate the design phase okay, when it comes time to fabricate I will need to drive out there and take measurements and double check everything.

I’d like to think in time that there will be an effective and highly accurate room scanning technology available (well, Leica has a great looking product, but at $8300 it is pricy), but with Canvas IO, the marketing promise far exceeds the actual performance. If you are just doing interior-designy stuff, where accuracy is relatively unimportant and you’re concerned with wall colors and how a particular piece of furniture might look in a space, etc., it would be fine I guess, but if the idea is to take the scan and design accurate components from it you would be heading for a world of hurt and pain.

All for this round. Thanks for visiting.

via Tumblr

No comments:

Post a Comment