After creating a tensorflow lite model that continuously performs bicycle helmet detection using quantized MobileNet SSD and Bicycle Helmet Dataset, it was made to be used in real time on Android by referring to TensorFlow Lite Object Detection Android Demo.
All models and code are ready. To run, download this repository, open it using Android Studio, and run it with an emulator or personal device.
After downloading the data set from kaggel, updating it on roboflow, visualizing the data, checking basic information, and dividing it into train, validation and test data at ratio of 70%, 20% and 10%.
check Roboflow-TFLite-Object-Detection.ipynb file, The file is very complex, so I recommend viewing it in colab.
As a result of learning, you can see the result that distinguishes between wearing a helmet and not wearing it. You can check out the demo video through the link below.
- TensorFlow Lite Object Detection Android Demo,https://github.com/tensorflow/examples/tree/master/lite/examples/object_detection/android
- How to Train a Custom TensorFlow Lite Object Detection Model, https://blog.roboflow.com/how-to-train-a-tensorflow-lite-object-detection-model/
- Bicycle Helmet Dataset, https://www.kaggle.com/andrewmvd/helmet-detection
- MobileNet SSD, https://github.com/tensorflow/models/tree/master/research/object_detection