As machine learning is getting bigger with the addition of new technologies, Google is set to help the developer community by offering an API on object detection. Called the TensorFlow Object Detection API, the open source solution was previously powering Google technologies such as NestCam, Image Search and Street View.
The first open source TensorFlow Object Detection API comes with a set of trainable models that enables object detection. Google has provided some parameters to train models that are on Microsoft’s Common Objects Context (COCO) data set to make the solution useful for researchers and developers. Additionally, there is a Jupyter notebook to perform out-of-the-box interference with the trainable detection models.
“This codebase is an open source framework built on top of TensorFlow that makes it easy to construct, train and deploy detection models. Our goal in designing this system was to support state-of-the-art models while allowing for rapid exploration and research,” Research Scientist Jonathan Huang and Software Engineer Vivek Rathod write in a blog post.
The Object Detection API is capable of localising and identifying multiple objects in a single image. It also works with MobileNet to deliver the same advanced experience on mobile devices that was initially limited to desktops.
Detecting objects in a sample image has become a trend in the machine learning space. While Google is accelerating TensorFlow with technologies like the latest Object Detection API, Facebook is offering solutions including Caffe2 and PyTorch and Microsoft has its separate Cognitive Toolkit.
Interestingly, all the solutions are advanced enough to detect visual objects automatically. But you need to pick one from the bouquet to enter the world of machine learning.
Google is providing a pre-trained model zoo alongside the TensorFlow Object Detection API to attract developer projects. Moreover, you can access the code from the GitHub repository to begin with your developments.