My react native tensorflow model gives the same prediction every result - reactjs

Question

I have trained a keras model with 6 classes for CNN Image Recognition. Then, I converted it to tensorflowjs using this command:
tensorflowjs_converter --quantization_bytes 1 --input_format=keras {saved_model_path} {target_path}
which then gives me 2 files (model.json, group1-shard1of1.bin). After adding it to my react native project and test the prediction function, it always returns the same result. I am converting my image to tensor base on this github thread:
const processImagePrediction = async (base64Image: any) => {
const fileUri = base64Image.uri;
const imgB64 = await FileSystem.readAsStringAsync(fileUri, {
encoding: FileSystem.EncodingType.Base64,
});
const imgBuffer = tf.util.encodeString(imgB64, 'base64').buffer;
const raw = new Uint8Array(imgBuffer);
let imageTensor = decodeJpeg(raw);
const scalar = tf.scalar(255);
imageTensor = tf.image.resizeNearestNeighbor(imageTensor, [224, 224]);
const tensorScaled = imageTensor.div(scalar);
//Reshaping my tensor base on my model size (224, 224)
const img = tf.reshape(tensorScaled, [1, 224, 224, 3]);
console.log("IMG: ", img);
const results = model.predict(img);
let pred = results.dataSync();
console.log("PREDICTION: ", pred);
}
I was expecting to have a different result but no matter what image I capture it always returns this result:
LOG IMG: {"dataId": {"id": 23}, "dtype": "float32", "id": 38, "isDisposedInternal": false, "kept": false, "rankType": "4", "scopeId": 6, "shape": [1, 224, 224, 3], "size": 150528, "strides": [150528, 672, 3]} LOG PREDICTION: [-7.7249579429626465, 4.73449182510376, 2.609705924987793, 20.0458927154541, -7.944214344024658, -18.101320266723633]
The numbers somehow change slightly but it retains the highest prediction (which is class number 4).
P.S. I'm new to image recognition so any tips would be appreciated.