UtkarshShivhare commited on
Commit
9b5c5c0
1 Parent(s): b3c024c

Upload app.py

Browse files
Files changed (1) hide show
  1. app.py +89 -0
app.py ADDED
@@ -0,0 +1,89 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ import tensorflow as tf
3
+ from PIL import Image
4
+ import numpy as np
5
+ import json
6
+ from tensorflow.keras.applications.vgg16 import VGG16,preprocess_input
7
+ from tensorflow.keras.preprocessing.image import img_to_array
8
+ from tensorflow.keras.preprocessing.text import Tokenizer,tokenizer_from_json
9
+ from tensorflow.keras.preprocessing.sequence import pad_sequences
10
+ from tensorflow.keras.models import Model
11
+
12
+ from keras.models import load_model
13
+
14
+ # Load the .h5 model
15
+ model = load_model('image_caption.h5')
16
+
17
+ with open('tokenizer_config.json', 'r') as f:
18
+ tokenizer_config = json.load(f)
19
+ tokenizer = tokenizer_from_json(tokenizer_config)
20
+ # tokenizer.word_index = eval(tokenizer_config)['word_index']
21
+
22
+ max_length=35
23
+ # Load pre-trained model
24
+ vgg_model = VGG16()
25
+ vgg_model = Model(inputs=vgg_model.inputs, outputs=vgg_model.layers[-2].output)
26
+
27
+ # Set Streamlit configurations
28
+ st.set_page_config(page_title="Image Captioning App", layout="wide")
29
+
30
+
31
+ # Function to preprocess the input image
32
+ def preprocess_image(image):
33
+ image = image.convert("RGB")
34
+ image = image.resize((224, 224))
35
+ image = img_to_array(image)
36
+ image = image.reshape((1, image.shape[0], image.shape[1], image.shape[2]))
37
+ image = preprocess_input(image)
38
+ return image
39
+
40
+ # Function to make predictions on the input image
41
+ def predict(image):
42
+ image = preprocess_image(image)
43
+ feature = vgg_model.predict(image, verbose=0)
44
+ preds = predict_caption(model, feature, tokenizer, max_length)
45
+ preds=preds[8:-7]
46
+ return preds
47
+
48
+ def idx_word(integer,tok):
49
+ for word,index in tok.word_index.items():
50
+ if index== integer:
51
+ return word
52
+ return None
53
+
54
+ def predict_caption(model,image,tok,max_len):
55
+ in_text="startseq"
56
+ for i in range(max_len):
57
+ seq=tok.texts_to_sequences([in_text])[0]
58
+ seq=pad_sequences([seq],max_len)
59
+ yhat = model.predict([image, seq], verbose=0)
60
+ yhat = np.argmax(yhat)
61
+ word = idx_word(yhat, tok)
62
+ if word is None:
63
+ break
64
+ in_text += " " + word
65
+ if word == 'endseq':
66
+ break
67
+ return in_text
68
+
69
+ # Streamlit app
70
+ def main():
71
+ st.title("Image Captioning App")
72
+ st.write("Upload an image and the app will predict its class.")
73
+
74
+ uploaded_image = st.file_uploader("Choose an image", type=["jpg", "jpeg", "png"])
75
+
76
+ if uploaded_image is not None:
77
+ image = Image.open(uploaded_image)
78
+ st.image(image, caption='Uploaded Image', use_column_width=True)
79
+ st.write("")
80
+
81
+ if st.button("Generate Caption"):
82
+ with st.spinner("Generating..."):
83
+ predictions = predict(image)
84
+
85
+ st.write(f"Top Caption:{predictions}")
86
+
87
+ # Run the app
88
+ if __name__ == "__main__":
89
+ main()