tflearn · vinayaksable2399 · Apr 12, 2019
diff --git a/examples/images/Fruits detection.py b/examples/images/Fruits detection.py
@@ -0,0 +1,96 @@
+
+""" 5 convolutional .
+
+
+    - [Fruit Dataset](http://www.site.uottawa.ca/~shervin/food/)
+"""
+import numpy as np         # dealing with arrays
+import os                  # dealing with directories
+from random import shuffle # mixing up or currently ordered data that might lead our network astray in training.
+import glob
+import cv2
+
+path = 'path of folder contains fruits folders'
+IMG_SIZE = 300 #image size for training
+LR = 1e-3 #learning rate
+
+MODEL_NAME = 'Fruits_dectector-{}-{}.model'.format(LR, '5conv-basic')
+no_of_fruits=7 # no. of folders contain in path file of fruits
+split=0.3 #spliting criteria
+no_of_images=100 # no. of image of each fruits
+
+def create_train_data(path):
+    training_data = []
+    folders=os.listdir(path)[0:no_of_fruits]
+    for i in range(len(folders)):
+        label = [0 for i in range(no_of_fruits)]
+        label[i] = 1
+        print(folders[i])
+        k=0
+        for j in glob.glob(path+"\\"+folders[i]+"\\*.jpg"):            
+            if(k==no_of_images):
+                break
+            k=k+1
+            img = cv2.imread(j)
+            img = cv2.resize(img, (IMG_SIZE,IMG_SIZE))
+            training_data.append([np.array(img),np.array(label)])
+    np.save('training_{}_{}_{}.npz'.format(no_of_fruits,no_of_images,IMG_SIZE),training_data)
+    shuffle(training_data)
+    return training_data,folders
+
+
+import tflearn
+from tflearn.layers.conv import conv_2d, max_pool_2d
+from tflearn.layers.core import input_data, dropout, fully_connected
+from tflearn.layers.estimator import regression
+
+
+#model building
+
+import tensorflow as tf
+tf.reset_default_graph()
+convnet = input_data(shape=[None, IMG_SIZE, IMG_SIZE, 3], name='input')
+
+convnet = conv_2d(convnet, 32, 5, activation='relu')
+
+convnet = max_pool_2d(convnet, 5)
+
+convnet = conv_2d(convnet, 64, 5, activation='relu')
+
+convnet = max_pool_2d(convnet, 5)
+
+convnet = conv_2d(convnet, 128, 5, activation='relu')
+convnet = max_pool_2d(convnet, 5)
+
+convnet = conv_2d(convnet, 64, 5, activation='relu')
+convnet = max_pool_2d(convnet, 5)
+
+
+convnet = conv_2d(convnet, 32, 5, activation='relu')
+convnet = max_pool_2d(convnet, 5)
+
+convnet = fully_connected(convnet, 1024, activation='relu')
+convnet = dropout(convnet, 0.8)
+
+convnet = fully_connected(convnet, no_of_fruits, activation='softmax')
+convnet = regression(convnet, optimizer='adam', learning_rate=LR, loss='categorical_crossentropy', name='targets')
+
+model = tflearn.DNN(convnet, tensorboard_dir='log')
+
+
+#data loading
+training_data,labels=create_train_data(path)
+a=int(len(training_data)*split)
+train = training_data[:-a]
+test=training_data[-a:]
+
+X = np.array([i[0] for i in train]).reshape(-1,IMG_SIZE,IMG_SIZE,3)
+Y = [i[1] for i in train]
+Y =np.array(Y)
+test_x = np.array([i[0] for i in test]).reshape(-1,IMG_SIZE,IMG_SIZE,3)
+test_y = [i[1] for i in test]
+
+model.fit({'input': X}, {'targets': Y}, n_epoch=10, validation_set=({'input': test_x}, {'targets': test_y}), 
+    snapshot_step=500, show_metric=True, run_id=MODEL_NAME)
+
+model.save(MODEL_NAME)