Commit
·
7f6f099
1
Parent(s):
13c4b7d
Upload 5 files
Browse files- app.py +52 -0
- feature_extractor.h5 +3 -0
- image_captioning_model.h5 +3 -0
- requirements.txt +2 -0
- tokenizer_data.pkl +3 -0
app.py
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import tensorflow as tf
|
| 2 |
+
from tensorflow.keras.utils import pad_sequences
|
| 3 |
+
import pickle
|
| 4 |
+
import streamlit as st
|
| 5 |
+
import numpy as np
|
| 6 |
+
from PIL import Image
|
| 7 |
+
from tensorflow.keras.applications.xception import preprocess_input
|
| 8 |
+
import logging
|
| 9 |
+
tf.get_logger().setLevel(logging.ERROR)
|
| 10 |
+
|
| 11 |
+
model = tf.keras.models.load_model('image_captioning_model.h5')
|
| 12 |
+
feature_extractor = tf.keras.models.load_model(r"feature_extractor.h5")
|
| 13 |
+
with open(r'tokenizer_data.pkl', 'rb') as f:
|
| 14 |
+
pickle_data = pickle.load(f)
|
| 15 |
+
tokenizer = pickle_data['tokenizer']
|
| 16 |
+
idx_to_word = pickle_data['word_mapping']
|
| 17 |
+
max_length = 35
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
def generate_caption(img):
|
| 21 |
+
in_text = 'startseq'
|
| 22 |
+
new_img = Image.open(img)
|
| 23 |
+
image = np.asarray(new_img.resize((299, 299)))
|
| 24 |
+
image = image.reshape((1, image.shape[0], image.shape[1], image.shape[2]))
|
| 25 |
+
image = preprocess_input(image)
|
| 26 |
+
feature = feature_extractor.predict(image, verbose=0)
|
| 27 |
+
for i in range(max_length):
|
| 28 |
+
sequence = tokenizer.texts_to_sequences([in_text])[0]
|
| 29 |
+
sequence = pad_sequences([sequence], max_length)
|
| 30 |
+
yhat = model.predict([feature, sequence], verbose=0)
|
| 31 |
+
yhat = np.argmax(yhat)
|
| 32 |
+
if yhat not in idx_to_word.keys():
|
| 33 |
+
break
|
| 34 |
+
else:
|
| 35 |
+
word = idx_to_word[yhat]
|
| 36 |
+
if word == 'endseq':
|
| 37 |
+
break
|
| 38 |
+
in_text += " " + word
|
| 39 |
+
|
| 40 |
+
return in_text.replace("startseq", '')
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
st.title("Image Caption Generator")
|
| 44 |
+
img = st.file_uploader("Upload image", type=["png", "jpg", "jpeg"],)
|
| 45 |
+
if img is not None:
|
| 46 |
+
st.image(Image.open(img), width=300)
|
| 47 |
+
|
| 48 |
+
if st.button("Generate Caption"):
|
| 49 |
+
if img is not None:
|
| 50 |
+
st.write(generate_caption(img))
|
| 51 |
+
else:
|
| 52 |
+
st.write("Please upload an image")
|
feature_extractor.h5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c6722667a1650c529d7399e825e13f2ed1fde48877a3b67509d75851c968cfd6
|
| 3 |
+
size 83863704
|
image_captioning_model.h5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d3886a34db43f251b5b19bb685a1b9e7ff5c0ba735c74ef7be12081d4ca16ef
|
| 3 |
+
size 65678548
|
requirements.txt
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
tensorflow
|
| 2 |
+
streamlit
|
tokenizer_data.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:46ed22b5947e48c1fc401ff55a336302389d90d173f9402201bcbafd5471f56c
|
| 3 |
+
size 401787
|