Yes. We thought about using car's value for quests/achievements. Haven't find reliable source for pricing data yet.
Current concept idea is very similar to one you suggested. Every car is assigned with some amount of points based on car's rarity. Player is then tasked to find certain models and get points for that.
We are thinking about adding 'social' element at some point: function to share car photos comment/react on them
We collected images, labeled the data and trained our model all on our own. We now have about 330k images (about 1800 models). We used resnet152 to train it.
We are just CS students and started this just for learning and experiencing with hands on computer vision and AI. So things are not very mature. Like we haven’t make the inferencing on the end user devices. The recognition runs on the server instead. And we don’t have many images as you do. We just crawled images on the web. Also, it doesn’t work very well with cars that can’t be found at where we lived, because we dont even know those models exist lol
We had/have this problem too. If nn "sees" unknown model it tries to classify it as one of known models. To avoid this, we added 'Other/Other' bucket and put ~3k images of different cars/buses/trucks in there. Now unknown models in most cases are classified as 'Other/Other' label.
I want to share a project me and couple more guys were working on last year - Car Hunters. Essentially we built a mobile game using Computer Vision and AI.
Some problems we had to solve while working on Car Hunters:
- we collected a pretty solid dataset of car images (~3mil photos so far). For that we built own software, which helps us to support car models hierarchy, maintain links, retrain recognition models and deliver recognition model-binaries to the end-user devices;
- make/model recognition is done on the end-user device and both android/iOS platforms run neural network inference on GPU;
- we built-in pretty fast and reliable object-tracker which is independent of platform as well.
So what is Car Hunters? We've all heard about Pokemon Go, where you hunt for virtual creatures. We offer you instead to get points and unlock achievements by searching for cars in real world!
More details and description can be found in Google Play and Apple App Store. Car Hunters is available only in 10 countries for now. Let me know if you unable to download it but would like to.
Hi jermaustin1!
Not exactly, we don't use users' data to extend dataset. Photos made by users are labeled (classified) by existing nn, and you don't want to use results produces by nn to retrain same nn.
Currently we have points, achievements for unique cars, some useful info, like prices of the cars. But we are constantly thinking what to add more.
What an insanely clever idea. Great work! 3M photos is a huge dataset; I’m guessing each photo already had associated metadata and you did not hand-classify? Where did you grab such a large dataset?
There are two main sources: used car sites and youtube. Used car sites have ads classified across make/model hierarchy already - you just need to filter away unsuitable photos - this task can be automated. And youtube is good for new models. We do some hand-classification as well, but only for rare cars, where images are not easy to find.
Current concept idea is very similar to one you suggested. Every car is assigned with some amount of points based on car's rarity. Player is then tasked to find certain models and get points for that.
We are thinking about adding 'social' element at some point: function to share car photos comment/react on them