![]() Use a rather small resolution, for example 640 x 480.Take/get at least 50 pictures per object.See the TensorFlow documentation for more details.įrom what I've read these are some best practices how to create the training data: Store the annotation files in the volume/data/annotations directory.ĭefine your objects in volume/data/label_map.pbtxt. ![]() Take pictures from the objects you want to recognize and replace the pictures in the volume/data/images directory. $ cp -R $/volume/data 2) Labelling of Images and Creation of TFRecords $ cd object-detection-anki-overdrive-cars In order to run the iOS app, a machine running MacOS is needed to compile TensorFlow 1.9.0: Setup of the iOS App with the trained Model This picture shows the iPad app which recognizes the two cars and the phone:Ĭheck out the documentation folder for more pictures and screenshots. To make the setup of the development environment as simple as possible, Docker containers are provided. The project also contains documentation how to train models to recognize other objects via TensorFlow Object Detection. ![]() This project contains a trained TensorFlow Object Detection model based on MobileNet and instructions how to run it in iOS apps and Jupyter notebooks. When an obstacle (phone) is detected, the cars stop. With an iOS app cars and phones can be detected on Anki Overdrive tracks. The object-detection-anki-overdrive-cars project includes a trained deep learning model to recognize items on Anki Overdrive tracks and it includes documentation how to train TensorFlow Object Detection models: TensorFlow Object Detection for Anki Overdrive Cars
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |