YOLOv9 model export as tflite model

[HumairaBano]

I am trying to export yolov9 model as tflite model but got error. Please guide me(I am using colab )

Load the YOLOv9 model

model = YOLO('/content/runs/detect/train2/weights/best.pt')

Export the model to TFLite format

model.export(format='tflite') # creates '.tflite'

and error
ONNX: export success ✅ 6.0s, saved as '/content/runs/detect/train2/weights/best.onnx' (96.9 MB)
TensorFlow SavedModel: starting TFLite export with onnx2tf 1.17.5...
TensorFlow SavedModel: export failure ❌ 26.7s: generic_type: cannot initialize type "StatusCode": an object with that name is already defined

ImportError Traceback (most recent call last)
in <cell line: 4>()
2 model = YOLO('/content/runs/detect/train2/weights/best.pt')
3 # Export the model to TFLite format
----> 4 model.export(format='tflite') # creates 'yolov8n_float32.tflite'

11 frames
/usr/local/lib/python3.10/dist-packages/tensorflow_lite_support/python/task/vision/core/tensor_image.py in
17
18 from tensorflow_lite_support.python.task.vision.core import color_space_type
---> 19 from tensorflow_lite_support.python.task.vision.core.pybinds import image_utils
20
21

ImportError: generic_type: cannot initialize type "StatusCode": an object with that name is already defined

[pderrenger]

Hello! It seems like you're facing a conflict issue during the TFLite export process, often due to package version mismatches.

A quick workaround to try in Colab is to restart the runtime and ensure you're using compatible versions of the packages involved in the conversion process (tensorflow, onnx, onnx-tf). Before exporting, you can attempt to explicitly upgrade or install specific versions of these packages and then try the export step again.

For instance:

!pip install tensorflow==2.6.0 onnx onnx-tf

After installing or upgrading, make sure to restart the Colab runtime:

Runtime -> Restart runtime

Then, attempt the export process again:

from ultralytics import YOLO

# Load the YOLOv9 model
model = YOLO('/content/runs/detect/train2/weights/best.pt')

# Export the model to TFLite format
model.export(format='tflite')

If issues persist, checking the specific versions of the packages involved for known compatibility issues could provide further insights. 🛠️

[glenn-jocher]

Hello! It seems like you're facing a conflict issue during the TFLite export process, often due to package version mismatches, especially when working on Colab which dynamically updates packages. A quick fix could be to refresh your Colab environment packages. Try the following steps:

1. Restart your Colab runtime to reset the environment: Runtime > Restart runtime.
2. Install a specific version of TensorFlow that's compatible with your needs, for example: !pip install tensorflow==2.6.0.
3. Ensure you have the Ultralytics YOLO package updated: !pip install -U ultralytics.
4. Re-run your model loading and exporting code.

from ultralytics import YOLO

# Load your YOLOv9 model
model = YOLO('/content/runs/detect/train2/weights/best.pt')

# Attempt to export again
model.export(format='tflite')

If issues persist, it could be helpful to share the exact version of TensorFlow installed and any additional export settings you might be using. Keeping a close eye on package versions can be critical in these situations. 🛠️

Let us know how it goes!

[Varun3713-creator]

hai @pderrenger ,
which tensorflow version is compatible with yolov9

implementation ("org.tensorflow:tensorflow-lite:2.14.0")
implementation ("org.tensorflow:tensorflow-lite-gpu:2.14.0")
implementation ("org.tensorflow:tensorflow-lite-support:0.4.4")
implementation ("org.tensorflow:tensorflow-lite-metadata:0.4.4")

I tried these but i am getting model load error.then i tried

implementation ("org.tensorflow:tensorflow-lite:2.16.2")
implementation ("org.tensorflow:tensorflow-lite-gpu:2.16.2")
implementation ("org.tensorflow:tensorflow-lite-support:0.4.4")
implementation ("org.tensorflow:tensorflow-lite-metadata:0.4.4")

Caused by: org.gradle.api.internal.artifacts.ivyservice.DefaultLenientConfiguration$ArtifactResolveException: Could not resolve all files for configuration ':app:debugRuntimeClasspath'.
which version can i use

[pderrenger]

Hello @Varun3713-creator,

Thank you for reaching out! It looks like you're encountering issues with exporting your YOLOv9 model to TFLite format and loading it with TensorFlow Lite. Let's address your concerns step-by-step.

Exporting YOLOv9 to TFLite

To export a YOLOv9 model to TFLite, you can use the following code snippet:

from ultralytics import YOLO

# Load the YOLOv9 model
model = YOLO('/content/runs/detect/train2/weights/best.pt')

# Export the model to TFLite format
model.export(format='tflite')  # creates 'best.tflite'

If you encounter errors during the export process, please ensure you are using the latest versions of the Ultralytics and TensorFlow packages. You can update them using:

pip install --upgrade ultralytics tensorflow

TensorFlow Lite Compatibility

Regarding the TensorFlow Lite versions, the following dependencies should generally work well:

implementation ("org.tensorflow:tensorflow-lite:2.16.2")
implementation ("org.tensorflow:tensorflow-lite-gpu:2.16.2")
implementation ("org.tensorflow:tensorflow-lite-support:0.4.4")
implementation ("org.tensorflow:tensorflow-lite-metadata:0.4.4")

However, compatibility issues can arise due to various reasons. If you are facing model load errors, please ensure that:

1. All dependencies are compatible: Sometimes, different versions of TensorFlow Lite components may not be fully compatible with each other. It's often best to use the same version for all TensorFlow Lite components.

2. Model Export: Ensure that the model export process completes successfully without errors. If there are any issues during export, they might propagate to the loading phase.

3. Environment: Verify that your development environment is correctly set up and that there are no conflicting packages.

Example Code for Loading TFLite Model

Here is an example of how you can load and run inference with a TFLite model in Python:

import tensorflow as tf

# Load the TFLite model
interpreter = tf.lite.Interpreter(model_path="best.tflite")
interpreter.allocate_tensors()

# Get input and output tensors
input_details = interpreter.get_input_details()
output_details = interpreter.get_output_details()

# Prepare your input data
input_data = ...  # Your input data here

# Set the tensor to point to the input data to be inferred
interpreter.set_tensor(input_details[0]['index'], input_data)

# Run the inference
interpreter.invoke()

# Get the results
output_data = interpreter.get_tensor(output_details[0]['index'])
print(output_data)

Further Assistance

If the issue persists, please provide a minimum reproducible example of your code and the exact error messages you are encountering. This will help us diagnose the problem more effectively. You can refer to our Minimum Reproducible Example Guide for more details.

For additional resources, you can visit our TFLite Integration Guide.

I hope this helps! If you have any further questions, feel free to ask. 😊