Spaces:
Build error
Build error
Commit
•
8596595
1
Parent(s):
c61020c
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
"""shadman_Image_classification.ipynb
|
3 |
+
|
4 |
+
Automatically generated by Colaboratory.
|
5 |
+
|
6 |
+
Original file is located at
|
7 |
+
https://colab.research.google.com/drive/1DCVsfqR-kOz3CHdB99IT75IHBzUtIf3M
|
8 |
+
"""
|
9 |
+
|
10 |
+
import tensorflow as tf
|
11 |
+
import PIL
|
12 |
+
import matplotlib.pyplot as plt
|
13 |
+
from tensorflow.keras import layers
|
14 |
+
import os
|
15 |
+
|
16 |
+
import pathlib
|
17 |
+
|
18 |
+
flower_dataset = "https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz"
|
19 |
+
dataset_path = tf.keras.utils.get_file('flower_photos',origin=flower_dataset,untar=True)
|
20 |
+
dataset_path = pathlib.Path(dataset_path)
|
21 |
+
|
22 |
+
roses = list(dataset_path.glob('roses/*'))
|
23 |
+
daisy = list(dataset_path.glob('daisy/*'))
|
24 |
+
print(roses[1])
|
25 |
+
|
26 |
+
PIL.Image.open(roses[10])
|
27 |
+
PIL.Image.open(daisy[10])
|
28 |
+
|
29 |
+
# training the dataset
|
30 |
+
|
31 |
+
training_images = tf.keras.preprocessing.image_dataset_from_directory(
|
32 |
+
dataset_path,
|
33 |
+
subset = "training",
|
34 |
+
validation_split = 0.25,
|
35 |
+
seed = 123,
|
36 |
+
image_size = (180, 180),
|
37 |
+
batch_size = 32
|
38 |
+
|
39 |
+
)
|
40 |
+
|
41 |
+
# validation of images
|
42 |
+
|
43 |
+
|
44 |
+
validation_images = tf.keras.preprocessing.image_dataset_from_directory(
|
45 |
+
dataset_path,
|
46 |
+
subset = "validation",
|
47 |
+
validation_split = 0.25,
|
48 |
+
seed = 123,
|
49 |
+
image_size = (180, 180),
|
50 |
+
batch_size = 32
|
51 |
+
|
52 |
+
)
|
53 |
+
|
54 |
+
flower_classes = training_images.class_names
|
55 |
+
print(flower_classes)
|
56 |
+
|
57 |
+
from tensorflow.python.framework.func_graph import flatten
|
58 |
+
# if there are 5 classes then
|
59 |
+
|
60 |
+
dataset_classes = 5
|
61 |
+
|
62 |
+
from tensorflow.keras.models import Sequential
|
63 |
+
|
64 |
+
model = Sequential([
|
65 |
+
#rescaling
|
66 |
+
layers.experimental.preprocessing.Rescaling(1./255, input_shape = (180,180,3)),
|
67 |
+
layers.Conv2D(16, 3 , padding='same' , activation='relu'),
|
68 |
+
layers.MaxPooling2D(),
|
69 |
+
layers.Conv2D(32, 3, padding ='same' , activation = 'relu'),
|
70 |
+
layers.MaxPooling2D(),
|
71 |
+
layers.Conv2D(64 , 3, padding='same', activation='relu'),
|
72 |
+
layers.MaxPooling2D(),
|
73 |
+
layers.Flatten(),
|
74 |
+
layers.Dense(128, activation='relu'),
|
75 |
+
layers.Dense(dataset_classes, activation='softmax')
|
76 |
+
])
|
77 |
+
|
78 |
+
model.compile(
|
79 |
+
optimizer='adam',
|
80 |
+
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
|
81 |
+
metrics=['accuracy'])
|
82 |
+
|
83 |
+
mymodel = model.fit(
|
84 |
+
training_images,
|
85 |
+
validation_data=validation_images,
|
86 |
+
epochs=10
|
87 |
+
)
|
88 |
+
|
89 |
+
def predict_input_image(img):
|
90 |
+
img_4d=img.reshape(-1,180,180,3)
|
91 |
+
prediction=model.predict(img_4d)[0]
|
92 |
+
return {flower_classes[i]: float(prediction[i]) for i in range(5)}
|
93 |
+
|
94 |
+
! pip install gradio
|
95 |
+
|
96 |
+
import gradio as gr
|
97 |
+
def predict_input_image(img):
|
98 |
+
img_4d=img.reshape(-1,180,180,3)
|
99 |
+
prediction=model.predict(img_4d)[0]
|
100 |
+
return {flower_classes[i]: float(prediction[i]) for i in range(5)}
|
101 |
+
|
102 |
+
image = gr.inputs.Image(shape=(180,180))
|
103 |
+
label = gr.outputs.Label(num_top_classes=5)
|
104 |
+
|
105 |
+
gr.Interface(fn=predict_input_image, inputs=image, outputs=label,interpretation='default').launch(debug='True')
|