The power of computer vision in your hand
iOS & Android
We have experience in developing image recognition apps for iOS as well as Android. Often our image recognition works with in C++ so that we can make use of the same source code in Objective C Kotlin or Java. This way you can easily integrate a powerful piece of image recognition in Xcode or Android studio. But we can also develop native computer vision software.
Smartphone developments
Nowadays smartphone cameras are capable enough to produce high quality images. Calculation power has had a massive increase as well. Finally the speed of data transfers has also increased with the rise of 4g, making it possible to process data in the cloud. This means that image recognition will be used in smartphone app more often and the market will become more mature.
Libraries
Studio diip has experience with developing computer vision libraries as a part of an app. We often deliver our software in libraruies that an app development company can add to the functionality of an app. Our responsibility usually starts with the receiving an image (photo or video) and ends with the communication of the detected results back to the app.
Deep learning image recognition
Many projects at Studio diip result in software that performs computer vision tasks with the use of deep learning. Often we train deep learning models with Keras and TensorFlow on our own systems or with the use of cloud computing. Eventually this gives us a model that can also be used in a smartphone application. For Android we reshape the model to a TensorFlow Lite model and for Apple devices we shape the model to a ML Core model. This way we make sure that we train a network independent from the platform, but we also make sure that both platforms are used in the most efficient way to detect. You can find an example of our deep learning capabilities on the post about the deep learning demo.
Offline or online recognition
Because smartphones have been increasing in computing power and the connection is so fast with the upcoming rise of 4g or even 5g networks, there are two possibilities that for the implementation of image recognition. It is possible to build an app in a way that computer vision will be performed on the phone itself, or that the image is uploaded to the server where it can be processed and the result can be sent back. Both ways have their pros and cons. On the smartphone the computing power is limited, but the user is not restricted to the connection and speed limits. On the other hand image recognition improvements should be provided through updates or the synchronization with data in the app. In the cloud on the other hand, there is plenty of computing power and updates can be done within seconds and sent to millions of users. The downside is that there should always be a constant stream of data to the phone of the user.
Offline or online recognition
Because smartphones have been increasing in computing power and the connection is so fast with the upcoming rise of 4g or even 5g networks, there are two possibilities that for the implementation of image recognition. It is possible to build an app in a way that computer vision will be performed on the phone itself, or that the image is uploaded to the server where it can be processed and the result can be sent back. Both ways have their pros and cons. On the smartphone the computing power is limited, but the user is not restricted to the connection and speed limits. On the other hand image recognition improvements should be provided through updates or the synchronization with data in the app. In the cloud on the other hand, there is plenty of computing power and updates can be done within seconds and sent to millions of users. The downside is that there should always be a constant stream of data to the phone of the user. For an example of an application like this see our project about 3D length measurement.
Automatic editing
With computer vision, it is not only possible to identify, count or show certain objects. It is also possible to adjust a photo or video stream live at the scene. This way objects can be replaced or removed from photos. Also the adjustments of colors, contrasts and other attributes of images or videos can be done with image recognition.