Fruits Classification using Deep Learning



Original Source Here

Fruits Classification using Deep Learning

Photo by Dragonlady on Dribbble

Task

In this project, we will be classifying a fruit and displaying its name as output from the given photo of the fruit as input.

Dataset- https://www.kaggle.com/sshikamaru/fruit-recognition

The dataset consists of 33 selected different kinds of fruits. Each folder is named after a fruit and contains over 400 images of that fruit in different angles and lightings. Based on the given image, we need to classify the fruit as one of the 33 categories.

Preprocessing –

  • First, import all the required libraries –
  • Download the data and unzip it to access the images and labels from your notebook.
  • There are total 33 types of fruits in our dataset. Map them and print them accordingly

output:

{'Pepper Green': 0, 'Lemon': 1, 'Cantaloupe': 2, 'Passion Fruit': 3, 'Pineapple': 4, 'Apricot': 5, 'Banana': 6, 'Pomegranate': 7, 'Pear': 8, 'Avocado': 9, 'Potato Red': 10, 'Plum': 11, 'Cucumber Ripe': 12, 'Strawberry': 13, 'Cactus fruit': 14, 'Raspberry': 15, 'Tomato': 16, 'Pepper Red': 17, 'Peach': 18, 'Blueberry': 19, 'Onion White': 20, 'Orange': 21, 'Watermelon': 22, 'Kiwi': 23, 'Limes': 24, 'Apple Granny Smith': 25, 'Apple Braeburn': 26, 'Cherry': 27, 'Grape Blue': 28, 'Corn': 29, 'Mango': 30, 'Clementine': 31, 'Papaya': 32}

Split the images into train, validation, test sets

  • Perform data augmentation by using ImageDataGenerator so that we can acquire more relevant data from the existing images by making minor alterations to the dataset.
  • Divide the training dataset into train set and validation set.

output:

Found 13309 images belonging to 33 classes.
Found 3314 images belonging to 33 classes.

Training the model

  • Compile and fit the model

output:

Epoch 1/10
415/415 [==============================] - 43s 103ms/step - loss: 2.1697 - accuracy: 0.3067 - val_loss: 0.8547 - val_accuracy: 0.6739
Epoch 2/10
415/415 [==============================] - 42s 102ms/step - loss: 0.7189 - accuracy: 0.7288 - val_loss: 0.3619 - val_accuracy: 0.8514
Epoch 3/10
415/415 [==============================] - 43s 103ms/step - loss: 0.3826 - accuracy: 0.8639 - val_loss: 0.2193 - val_accuracy: 0.9215
Epoch 4/10
415/415 [==============================] - 43s 102ms/step - loss: 0.2220 - accuracy: 0.9222 - val_loss: 0.1437 - val_accuracy: 0.9478
Epoch 5/10
415/415 [==============================] - 43s 104ms/step - loss: 0.2096 - accuracy: 0.9319 - val_loss: 0.0678 - val_accuracy: 0.9789
Epoch 6/10
415/415 [==============================] - 43s 103ms/step - loss: 0.1355 - accuracy: 0.9560 - val_loss: 0.0984 - val_accuracy: 0.9635
Epoch 7/10
415/415 [==============================] - 44s 106ms/step - loss: 0.1193 - accuracy: 0.9596 - val_loss: 0.0309 - val_accuracy: 0.9900
Epoch 8/10
415/415 [==============================] - 43s 103ms/step - loss: 0.1169 - accuracy: 0.9626 - val_loss: 0.0714 - val_accuracy: 0.9740
Epoch 9/10
415/415 [==============================] - 43s 104ms/step - loss: 0.1030 - accuracy: 0.9677 - val_loss: 0.0205 - val_accuracy: 0.9934
Epoch 10/10
415/415 [==============================] - 42s 102ms/step - loss: 0.0963 - accuracy: 0.9704 - val_loss: 0.0706 - val_accuracy: 0.9749

Loss and accuracy graphs –

Predicting a fruit from the test set

  • I have chosen a random image from the test set
  • Load the image and convert it to appropriate size

output:

Prediction is Pineapple.

Conclusion

Hence, we have trained a sequential model in keras to predict the name of the fruit with an image of a fruit as the input.

Notebook link: https://cainvas.ai-tech.systems/use-cases/fruits-classification-app/

Credit: Amruta Koshe

AI/ML

Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: