This module implements the rdk:service:mlmodel
API in an tflite_cpu
model.
With this model, you can deploy TensorFlow Lite model files as long as your models adhere to the model requirements.
It is supported on any CPU and Linux, Raspbian and MacOS machines.
To work with the tflite_cpu
ML model service, an ML model is comprised of a .tflite model file which defines the model, and optionally a .txt labels file which provides the text labels for your model.
With the tflite_cpu
ML model service, you can deploy:
- a model from the registry
- a model trained outside the Viam platform that you have uploaded
- a model available on your machine
Navigate to the CONFIGURE tab of your machine in the Viam app.
Add ML model / TFLite CPU
to your machine.
Enter a name or use the suggested name for your service.
Click Add module and click Create.
You can choose to configure your service with an existing model on the machine or deploy a model onto your machine. To do so in BUILDER mode on the Viam app:
- To configure your service and deploy a model onto your machine, select Deploy model on machine for the Deployment field in the resulting ML model service configuration pane.
- Click Select model. In the modal that appears, search for models from your organization or the Registry. You can select a model to see more details about it, and then select the model to deploy it to your machine.
- Also, optionally select the Number of threads.
- Click Save at the top right of the window to save your changes.
- To configure your service with an existing model on the machine, select Path to existing model on machine for the Deployment field.
- Then specify the absolute Model path and any Optional settings such as the absolute Label path and the Number of threads.
You can also chose to use JSON mode in the Viam app. Add the tflite_cpu
ML model object to the services array in your JSON configuration:
"services": [
{
"name": "<mlmodel_name>",
"type": "mlmodel",
"model": "tflite_cpu",
"attributes": {
"model_path": "${packages.<model_name>}/<model-name>.tflite",
"label_path": "${packages.<model_name>}/labels.txt",
"num_threads": <number>
}
},
... // Other services
]
{
"packages": [
{
"package": "39c34811-9999-4fff-bd91-26a0e4e90644/my_fruit_model",
"version": "YYYY-MM-DDThh-mm-ss",
"name": "my_fruit_model",
"type": "ml_model"
}
], ... // < Insert "components", "modules" etc. >
"services": [
{
"name": "fruit_classifier",
"type": "mlmodel",
"model": "tflite_cpu",
"attributes": {
"model_path": "${packages.my_fruit_model}/my_fruit_model.tflite",
"label_path": "${packages.my_fruit_model}/labels.txt",
"num_threads": 1
}
}
]
}
The "packages"
array shown above is automatically created when you deploy the model.
You do not need to edit the configuration yourself, except if you wish to specify a specific version for your deployed model.
The following parameters are available for a "tflite_cpu"
model:
Parameter | Required? | Description |
---|---|---|
model_path |
Required | The absolute path to the .tflite model file, as a string . |
label_path |
Optional | The absolute path to a .txt file that holds class labels for your TFLite model, as a string . This text file should contain an ordered listing of class labels. Without this file, classes will read as "1", "2", and so on. |
num_threads |
Optional | An integer that defines how many CPU threads to use to run inference. Default: 1 . |
If you Deploy model on machine, model_path
and label_path
will be automatically configured in the format "${packages.<model_name>}/<model-name>.tflite"
and "${packages.<model_name>}/labels.txt"
respectively.
If you take the Path to existing model on machine approach, your model and label paths do not have to be in the same format.
For example, they might resemble home/models/fruit/my_fruit_model.tflite
.
Save the configuration.
Models trained in the Viam app meet these requirements by design.
We strongly recommend that you package your TensorFlow Lite model with metadata in the standard form.
In the absence of metadata, your tflite_cpu
model must satisfy the following requirements:
- A single input tensor representing the image of type UInt8 (expecting values from 0 to 255) or Float 32 (values from -1 to 1).
- At least 3 output tensors (the rest won’t be read) containing the bounding boxes, class labels, and confidence scores (in that order).
- Bounding box output tensor must be ordered [x x y y], where x is an x-boundary (xmin or xmax) of the bounding box and the same is true for y. Each value should be between 0 and 1, designating the percentage of the image at which the boundary can be found.
These requirements are satisfied by a few publicly available model architectures including EfficientDet, MobileNet, and SSD MobileNet V1. You can use one of these architectures or build your own.
The ML model service only runs your model on the machine. To use the inferences from the model, you must use an additional service such as a vision service.
For example: