Running .tflite model on android

General Discussion
,
#1

Hello, I have a real time yolo model. My yolo model detects dinners in the image. Now I want food to be classified by type Ex: image-> evening food(yolo)-> soup(classification). My question : I did train image classification model, but I don’t know how to run it after the yolo model. I would love any help. Thanks…

#3

Hi @Laura1

Do you want to use it with python, java or swift api?

#5

Yes, @George_Soloupis

Additional info:

I used here for yolo. https://github.com/AarohiSingla/TFLite-Object-
Detection-Android-App-Tutorial-Using-YOLOv5

i will use more than one yolo and classification model.

Foods in tray. First there is a yolo for the tray. Then the tray classification is eveningTray or breakfastTray(vgg16 or efficient net image classification). Then, if the tray is in the evening, the evening yolo will work for the detection of food. Then reclassification in recognition. Main course,soup,salad,dessert.

#7

Hello @George_Soloupis ,

I want to ask one more thing
Should there be more than one .tflite models on android?
My models
Tray Yolo model (tray.tflite)
Dinner Breakfast Image Classification model VGG16 (DinnerBreakfast.tflite)
Dinner food Yolo (dinnerfood.tflite)
Breakfast food Yolo (breakfast.tflite)
SoupMainCourseSaladDessert Image Classification VGG16 (SoupMainCourseSaladDessert.tflite)
TeaOliviesBreadCheese Image Classification Model VGG16 (teaoliviesbreadcheeese.tflite)

func TrayYolo()
func TrayClassification()
if(tray == evening) func dinnerYolo() func SoupMainCourseSaladDessert()
else func breakfastYolo() func TeaOliviesBreadCheese()

Should I combine models trained in python in androd like this?

#8

There can be multiple tflite models inside an android application. You have to close() the TensorFlow Lite Interpreter when you do not need it anymore.

Regards

#10

Hi @George_Soloupis
I need to cut objects detected with yolo and pass them through the classification model. Please suggest me source.

TFLite-Object-Detection-Android-App-Tutorial-Using-YOLOv5/YoloV5Classifier.java at main · AarohiSingla/TFLite-Object-Detection-Android-App-Tutorial-Using-YOLOv5 · GitHub Here is the box part, I did stuck here.