import deeptrack as dt

import numpy as np

import random

import os

samples = []

data_folder = "training_data\\big_particles"

for sample in os.listdir(data_folder):

samples.append(dt.LoadImage(os.path.join(data_folder, sample))()._value[:, :, :3] / 256)

model = dt.models.LodeSTAR(input_shape=(None, None, 3))

model.load_weights("models\\big_particles\\weights")

train_set = (

dt.Value(lambda: random.choice(samples))

>> dt.Add(lambda: np.random.randn() * 0.1)

>> dt.Gaussian(sigma=lambda:np.random.uniform(0, 0.2))

>> dt.Multiply(lambda: np.random.uniform(0.6, 1.2))

)

model.fit(

train_set,

epochs=40,

batch_size=8,

)

model.save_weights("models\\big_particles\\weights")