Yolov8 TFLite Model on Android

Laura1 · December 10, 2023, 11:09am

Hello everyone , I’m trying to run yolov8 model on android. This is the Yolov5Classifier.java file. It works for yolov5 model.Where do I need to change in Yolov5Classifier.java file for yolov8 ? Which parts should I pay attention to ?
I added the YOLOV5Classifier.java file here.

github.com

usertttwm/YoloV5Android/blob/main/YOLOV5Classifier.java

public class YoloV5Classifier implements Classifier {

    /**
     * Initializes a native TensorFlow session for classifying images.
     *
     * @param assetManager  The asset manager to be used to load assets.
     * @param modelFilename The filepath of the model GraphDef protocol buffer.
     * @param labelFilename The filepath of label file for classes.
     * @param isQuantized   Boolean representing model is quantized or not
     */
    public static YoloV5Classifier create(
            final AssetManager assetManager,
            final String modelFilename,
            final String labelFilename,
            final boolean isQuantized,
            final int inputSize
            /*final int[] output_width,
            final int[][] masks,
            final int[] anchors*/)
            throws IOException {

This file has been truncated. show original

Thanks.

BadarJaffer · December 11, 2023, 12:18pm

@Laura1

1.Check the create method where the model is initialized. Ensure that you are using the correct filepath for the YOLOv8 model GraphDef protocol buffer (modelFilename).

2.Confirm that the input size parameter (inputSize) matches the expected input size for the YOLOv8 model. YOLO models often have specific input size requirements.

Examine how the output of the YOLOv5 model is processed in the code. The YOLOv8 model might have differences in terms of the number of bounding box predictions, classes, or other output details.
If YOLOv8 uses a different set of anchors or masks, you might need to update the corresponding variables in the code.

I’d recommend referring to documentation for further details: Home - Ultralytics YOLOv8 Docs

Laura1 · December 11, 2023, 4:31pm

Hi @BadarJaffer . Thank you very much for reply. Yolov5 model input shape (1, 3, 416, 416) output shape (1, 10647, 11)
Yolov8 model input shape (1, 3, 416, 416) and output shape(s) (1, 6, 3549)

/final int[] output_width,
final int[][] masks,
final int[] anchors/) These values are not used. But yolov5 works