Very interesting. I often kill time browsing through photos on houzz.com. There are also user discussion forums (formerly gardenweb.com I think) that talk about all sorts of home topics like house design, plumbers, etc.
I then think of computer build discussions[1] for gaming or AutoCAD work that insert links from pcpartpicker.com. Notice how pcpartpicker urls enables posters to easily dissect each others' "what if?" component choices.
It seems like a website could really combine the enthusiasm of the houzz/pinterest crowd with the deeper integration of a pcpartpicker.com. Instead of the "green tags"[2] on houzz being someone else's choice (and hence an unchangeable "ad"), the green tags would be picked by the user of a particular 3D generated model. The "green tags" could also be choices crowdsourced from others including freelance interior designers.
The service works like this: The homeowner takes several pictures of the space they want to decorate — ideally a photo from each of the room’s corners. Then, these photos are uploaded to Modsy after which the homeowner takes a style quiz about their likes and dislikes. This takes about 5 minutes and basically involves clicking on photos. Then Modsy takes over. About 48 hours later, the results are delivered.
Modsy uses a technique called point cloud reconstruction. This is the easy part, Tellerman says explaining the whole process is similar to a film pipeline with a preproduction, production and post production. The homeowner provides the service with the necessary media. The company then puts it through its rendering systems that reconstructions the room in a 3D space and then a system plugs in furnishings per the homeowner’s style direction.
What what?? I thought 2d to 3d reconstruction was an ongoing computer vision problem? This talks like it is solved.
Just getting a "floor plan" or empty room is much easier. But yes, it's an active area of research in Computer Vision / Graphics. Later on in the article it says "90% automated" and that things are checked. That'll be good enough to make this cost feasible.
Disclaimer: I'm related to one of the founders (but I don't work there).
How are we sure it's actually automated? I feel like they've just hired some kid to sit on autocad and recreate rooms based on the couple of pictures someone has taken.
It's a cool idea but seems like it will be difficult to get paying customers at $40 a pop. The value is there, but people are not used to paying this way.
I wonder if instead revenue could be extracted from purchases made based on the renderings, or via other partnerships.
Yeah, seems like bringing the showroom into your house would be a great way to sell furniture and accessories. Make this a platform for retailers and manufacturers!
This is coming just in time. I moved into my new place nearly 6 months ago (more than double the square footage) and the place is still sitting mostly empty because I haven't had the time to do anything about it. Really interested to try this out.
Killer app for VR+interior decorators. Regular people may not pay for a VR headset, but bringing this gear to help you visualize the work would be huge.
Even better, for architects doing new home construction.
I tried to create an MVP in a similar vein a while back, except instead of doing the whole 3D rendering route, I tested with Photoshoppers from ODesk with instructions to select products from a few well-known sellers of furniture/decor.
What caused me to give up at the time was in my initial tests with friends and family I found that the big problem is the cost of iterations quickly outpaces the price paid if the customer is not happy with what is presented. Often people have no idea what they want--they just can say "yeah, this looks nice, but I'll know it when I see it."
Even if they have a large library of pre-rendered products, there is still the cost of the time for the designer to pick things that work and place them. In theory you might be able to get this down low enough with machine learning to suggest products based on other happy customers with similar preferences and I'm curious if that is the direction you're leaning in to have fully automated recommendations.
There are other competing services out there that provide much less cool "mood board" style selections, and that is their attempt to get the price low by not wasting any time with trying to show it in context with the room. They also seem to charge ~$300.
Would love to know how you guys are addressing the issue of iterations as I didn't see that mentioned anywhere. Do you have a "happy customer guarantee?" If I pay $30-$40 and receive designs I'm really unhappy with, how do you guys handle that?
Congrats on what looks like an awesome product with some serious technology behind it. If this works as presented it will blow the competition out of the water, and if you can come up with a way to reduce the cost sufficiently, I think a lot of people would be very interested in getting great design suggestions they can easily visualize for a very low price.
Signed-up for the beta and will be filling out the style quiz with my wife tonight. I would LOVE to be bumped up in the list to help beta this as we are both super stressed about how to fill several rooms in our home and fill an empty wall in our living room. There's only so many crappy Pinterest publisher-spam links I can click on or insanely overpriced purchasable products on Houzz I can pass over before I want to throw my iPad against the wall.
[+] [-] jasode|10 years ago|reply
I then think of computer build discussions[1] for gaming or AutoCAD work that insert links from pcpartpicker.com. Notice how pcpartpicker urls enables posters to easily dissect each others' "what if?" component choices.
It seems like a website could really combine the enthusiasm of the houzz/pinterest crowd with the deeper integration of a pcpartpicker.com. Instead of the "green tags"[2] on houzz being someone else's choice (and hence an unchangeable "ad"), the green tags would be picked by the user of a particular 3D generated model. The "green tags" could also be choices crowdsourced from others including freelance interior designers.
[1]http://www.tomshardware.com/answers/id-1988274/good-computer...
[2]example: http://www.houzz.com/photos/30526217/Black-Is-The-New-Farmho...
[+] [-] hacker_9|10 years ago|reply
Modsy uses a technique called point cloud reconstruction. This is the easy part, Tellerman says explaining the whole process is similar to a film pipeline with a preproduction, production and post production. The homeowner provides the service with the necessary media. The company then puts it through its rendering systems that reconstructions the room in a 3D space and then a system plugs in furnishings per the homeowner’s style direction.
What what?? I thought 2d to 3d reconstruction was an ongoing computer vision problem? This talks like it is solved.
[+] [-] boulos|10 years ago|reply
Disclaimer: I'm related to one of the founders (but I don't work there).
[+] [-] mbratkova|10 years ago|reply
[+] [-] seangrant|10 years ago|reply
[+] [-] jkestner|10 years ago|reply
[+] [-] WhitneyLand|10 years ago|reply
I wonder if instead revenue could be extracted from purchases made based on the renderings, or via other partnerships.
[+] [-] jkestner|10 years ago|reply
[+] [-] james33|10 years ago|reply
[+] [-] darkvertex|10 years ago|reply
[+] [-] jkestner|10 years ago|reply
Even better, for architects doing new home construction.
[+] [-] shostack|10 years ago|reply
What caused me to give up at the time was in my initial tests with friends and family I found that the big problem is the cost of iterations quickly outpaces the price paid if the customer is not happy with what is presented. Often people have no idea what they want--they just can say "yeah, this looks nice, but I'll know it when I see it."
Even if they have a large library of pre-rendered products, there is still the cost of the time for the designer to pick things that work and place them. In theory you might be able to get this down low enough with machine learning to suggest products based on other happy customers with similar preferences and I'm curious if that is the direction you're leaning in to have fully automated recommendations.
There are other competing services out there that provide much less cool "mood board" style selections, and that is their attempt to get the price low by not wasting any time with trying to show it in context with the room. They also seem to charge ~$300.
Would love to know how you guys are addressing the issue of iterations as I didn't see that mentioned anywhere. Do you have a "happy customer guarantee?" If I pay $30-$40 and receive designs I'm really unhappy with, how do you guys handle that?
Congrats on what looks like an awesome product with some serious technology behind it. If this works as presented it will blow the competition out of the water, and if you can come up with a way to reduce the cost sufficiently, I think a lot of people would be very interested in getting great design suggestions they can easily visualize for a very low price.
Signed-up for the beta and will be filling out the style quiz with my wife tonight. I would LOVE to be bumped up in the list to help beta this as we are both super stressed about how to fill several rooms in our home and fill an empty wall in our living room. There's only so many crappy Pinterest publisher-spam links I can click on or insanely overpriced purchasable products on Houzz I can pass over before I want to throw my iPad against the wall.
Congrats!
[+] [-] brianbarker|10 years ago|reply
[+] [-] boulos|10 years ago|reply