Im trying to classify mushrooms using this dataset.
Im using tf.keras.utils.image_dataset_from_directory to import my dataset
How I handle the dataset
DATADIR = "/content/drive/MyDrive/Mushrooms/dataset"
data_dir = pathlib.Path(DATADIR).with_suffix('')
batch_size = 32
img_height = 200
img_width = 200
training_ds = tf.keras.utils.image_dataset_from_directory(
data_dir,
validation_split = 0.2,
subset = "training",
seed = 555,
image_size = (img_height, img_width),
batch_size = batch_size
)
validation_ds = tf.keras.utils.image_dataset_from_directory(
data_dir,
validation_split = 0.2,
subset = "validation",
seed = 555,
image_size = (img_height, img_width),
batch_size = batch_size
)
class_names = training_ds.class_names
print(class_names)
image_shape = []
for image_batch, labels_batch in training_ds:
image_shape = image_batch.shape
print(image_batch.shape)
print(labels_batch.shape)
break
print(image_shape)
AUTOTUNE = tf.data.AUTOTUNE
train_ds = training_ds.cache().prefetch(buffer_size = AUTOTUNE)
validation_ds = validation_ds.cache().prefetch(buffer_size = AUTOTUNE)
How I train the model
from tensorflow.keras.layers import Rescaling, Conv2D , MaxPooling2D , Flatten, Dense, Dropout
model = tf.keras.Sequential([
Rescaling(1./255),
Conv2D(256,3,activation = 'relu'),
MaxPooling2D(pool_size =(2,2)),
Conv2D(128,3,activation = 'relu'),
MaxPooling2D(pool_size =(2,2)),
Conv2D(64,3,activation = 'relu'),
MaxPooling2D(pool_size =(2,2)),
Flatten(),
Dense(128, activation='relu'),
Dropout(0.5),
Dense(number_of_classes,activation = 'softmax')
])
model.build(input_shape = (image_shape[0],image_shape[1],image_shape[2],image_shape[3]))
model.compile(
optimizer = 'adam',
loss = tf.keras.losses.SparseCategoricalCrossentropy(),
metrics=['sparse_categorical_accuracy']
)
model.summary()
model.fit(training_ds, validation_data=validation_ds,epochs = 3,use_multiprocessing = True)
I am currently getting the error
InvalidArgumentError: Graph execution error:
2 root error(s) found.
(0) INVALID_ARGUMENT: jpeg::Uncompress failed. Invalid JPEG data or crop window.
[[{{node decode_image/DecodeImage}}]]
[[IteratorGetNext]]
[[IteratorGetNext/_4]]
(1) INVALID_ARGUMENT: jpeg::Uncompress failed. Invalid JPEG data or crop window.
[[{{node decode_image/DecodeImage}}]]
[[IteratorGetNext]]
0 successful operations.
0 derived errors ignored. [Op:__inference_train_function_3271]
I have tried reinstalling the dataset, since I thought the dataset was corrupt. I think it is some sort of resizing issue. I have got this dataset working by converting the images to a numpy array, but I wanted to try work with the images directly.
