Detect objects from image using Tensorflow 2 SavedModel.
Run the following command in your Node-RED user directory - typically ~/.node-red
npm install Lapland-UAS-Tequ/tequ-node-red-object-detection-tf2-sm
Run prediction on input image using Tensorflow 2 Savedmodel.
Input image must be image buffer in 'msg.payload'.
Outputs:
- Prediction result
- Metagraph (tf.node.getMetaGraphsFromSavedModel)
- Tensorflow memory status (tf.memory())
Excepts following content in model folder:
- saved_model.pb (Tensorflow 2 saved model)
- variables (folder that might include files for saved model)
- labels.json (array of label strings that model can detect)
- metadata.json (metadata of used model in JSON-format, optional)
Supported image formats:
- JPG, PNG, BMP, GIF
Supports:
- Saved model trained using tequ-tf2-ca-training-pipeline
- TensorFlow 2 Detection Model Zoo models:
- SSD MobileNet v2 320x320
- SSD MobileNet V2 FPNLite 320x320
- SSD MobileNet V2 FPNLite 640x640
Others might work too, but have not been tested.
To train a model, please look:
https://github.com/Lapland-UAS-Tequ/tequ-tf2-ca-training-pipeline
Dependencies: