关于深度学习的正确率一直不提升

import numpy as np
import tensorflow as tf
from tensorflow.keras import Model,Sequential
from tensorflow.keras.layers import Conv2D,MaxPooling2D,Flatten,Dense,Dropout
from tensorflow.keras.applications.vgg16 import VGG16
from sklearn.model_selection import train_test_split
from keras_preprocessing.image import ImageDataGenerator
import os.path
import glob
import random

image_path=glob.glob('gar/*/*.jpg')

label_type=[image_p.split('\\')[1] for image_p in image_path]
labels=np.unique(label_type)

index_to_label=dict((l,n)for l,n in enumerate(labels))
label_to_index=dict((n,l)for l,n in index_to_label.items())


all_labels=[label_to_index.get(name)for name in label_type]


random_index=np.random.permutation(len(image_path))

image_path=np.array(image_path)[random_index]
all_labels=np.array(all_labels)[random_index]


sep=int(len(image_path)*0.7)
train_image_path=image_path[:sep]
train_y=all_labels[:sep]
test_image_path=image_path[sep:]
test_y=all_labels[sep:]

train=tf.data.Dataset.from_tensor_slices((train_image_path,train_y))
test=tf.data.Dataset.from_tensor_slices((test_image_path,test_y))

def load_pic(path,label):
    image=tf.io.read_file(path)
    image=tf.image.decode_jpeg(image,channels=3)
    image=tf.image.resize(image,[256,256])
    image=tf.cast(image,tf.float32)
    image=image/255
    return image,label
train=train.map(load_pic)
train=train.shuffle(1000).batch(64)
test=test.map(load_pic)
test=test.shuffle(1000).batch(64)
def get_model():
    gmodel=Sequential([

        Conv2D(64, (3, 3), activation='relu', input_shape=(256,256,3)),
        Conv2D(64,(3,3),activation='relu'),
        MaxPooling2D((2,2),strides=2,padding='same'),
        Conv2D(128, (3, 3), activation='relu'),
        Conv2D(128, (3, 3), activation='relu'),
        MaxPooling2D((2,2), strides=2, padding='same'),
        Conv2D(256, (3, 3), activation='relu'),
        Conv2D(256, (3, 3), activation='relu'),
        MaxPooling2D((2,2), strides=2, padding='same'),
        Conv2D(512, (3, 3), activation='relu'),
        Conv2D(512, (3, 3), activation='relu'),
        MaxPooling2D((2,2), strides=2, padding='same'),
        Conv2D(512, (3, 3), activation='relu'),
        Conv2D(512, (3, 3), activation='relu'),
        MaxPooling2D((2,2), strides=2, padding='same'),
        Flatten(),
        Dense(256),
        Dense(4, activation='softmax')

    ])


    return gmodel
model=get_model()


model.compile(
    optimizer=tf.keras.optimizers.Adam(0.0001),
    loss="categorical_crossentropy",
    metrics=['accuracy']
)
history=model.fit(train,epochs=5,validation_data=test)

为什么我这个代码在深度学习的时候正确率一直在20%多并且loss上升。我希望给图片分为四类。我尝试过换模型，optimizer和loss function但是都没有用。环境是python3.7，tensorflow2.0，gpu是2070

很多神经元死掉了吧