It’s been a while since I engaged here. We’ve been really busy with this AI thing. My name is Gabriel Eze and I’m part of the people at Touchabl Pictures. We are a technology startup and our area of focus is object recognition > computer vision > deep tech > AI.
We just released the iOS version of this android app in public beta. It’s a simple, handy fashion recognition app (works with fashion items only).
I’ve tested it with 3 iPhones but it seem to work in 2 and not work in the 3rd one. This is not the case for the android version. Both apps uses the same API we wrote to communicate with our AI for fashion recognition. I’ve asked almost everyone in my phonebook but nearly all of them don’t have iPhone; makes me wonder if I need new friends (pun intended).
Please, if you have an iPhone or iPad, would you be kind to help install it and share screenshot?
(whether it works or not)
** The app will ask you for permission to use your device camera, kindly grant it the permission as it currently get image input from the camera only. I just need to be sure if it’s a device thing, network performance issue, or something else. Thanks you in advance.
A lady came on Shark Tank about 2 years ago with similar idea (FashionTap) but never raised any money. She was grilled by the sharks. I think you should watch her Pitch and see the comments from the sharks.
Having said that, IMO, I think it will cost you time and money to make this a profitable adventure. You will not only be building an algorithm that detects stuffs but also building a social network (for people to post stuffs - chicken egg situation). Sadly, the market might just not be big enough for the efforts.
Advice - find a way to build an app that can easily be plugged into facebook, instagram, snapchat, or twitter…so once this app is installed on phone, anyone with these social networks can quickly look up items from friends pictures or images.
Thanks for the feedback. If you don’t mind, I would like to use your test images as training data for our next round of training for the model. Is it okay by you? If you have more photos you used to test, it will be very helpful to us as well.
P.S
I’m beginning to realize that pictures taken in Nigeria tend to have a grey colour overlay. In addition, most of our training data are “oyibo” skin coloured. We have a filter that ignores skin colours during the colour recognition so it can narrow down to the object colour; unfortunately, it was optimized for caucasian skin only (something we must correct). Your test image (and other’s) will help us further train the model to recognize and ignore black skin as well during object and colour detections respectively.
