Update app.py
Browse files
app.py
CHANGED
@@ -4,15 +4,59 @@ import streamlit as st
|
|
4 |
import requests
|
5 |
from transformers import pipeline
|
6 |
from PIL import Image
|
7 |
-
from googletrans import Translator
|
8 |
import sqlite3
|
9 |
|
10 |
# Initialize Streamlit app
|
11 |
st.set_page_config(page_title="Image Caption Generator", layout="wide")
|
12 |
|
13 |
-
|
14 |
|
15 |
@st.cache_resource
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
# Image to Text
|
17 |
def image_to_text(url):
|
18 |
# Load a transformer
|
@@ -101,9 +145,7 @@ def login_section():
|
|
101 |
except sqlite3.OperationalError as e:
|
102 |
st.error(f"An error occurred while trying to log in: {e}")
|
103 |
|
104 |
-
|
105 |
-
translated = translator.translate(caption, dest=target_language)
|
106 |
-
return translated.text
|
107 |
|
108 |
def predict(cap_col, target_language):
|
109 |
captions = []
|
@@ -111,13 +153,13 @@ def predict(cap_col, target_language):
|
|
111 |
pred_caption = image_to_text(url)
|
112 |
|
113 |
cap_col.markdown('#### Predicted Captions:')
|
114 |
-
translated_caption =
|
115 |
captions.append(translated_caption)
|
116 |
|
117 |
for _ in range(4):
|
118 |
pred_caption = image_to_text(url)
|
119 |
if pred_caption not in captions:
|
120 |
-
translated_caption =
|
121 |
captions.append(translated_caption)
|
122 |
|
123 |
cap_col.markdown('<div class="caption-container">', unsafe_allow_html=True)
|
@@ -136,7 +178,9 @@ def generate_caption_section():
|
|
136 |
img_upload = st.file_uploader(label='Upload Image', type=['jpg', 'png', 'jpeg'])
|
137 |
|
138 |
# Language selection dropdown
|
139 |
-
|
|
|
|
|
140 |
|
141 |
# Process image and generate captions
|
142 |
if img_url:
|
|
|
4 |
import requests
|
5 |
from transformers import pipeline
|
6 |
from PIL import Image
|
|
|
7 |
import sqlite3
|
8 |
|
9 |
# Initialize Streamlit app
|
10 |
st.set_page_config(page_title="Image Caption Generator", layout="wide")
|
11 |
|
12 |
+
|
13 |
|
14 |
@st.cache_resource
|
15 |
+
|
16 |
+
##Bloom
|
17 |
+
API_URL = "https://api-inference.huggingface.co/models/bigscience/bloom"
|
18 |
+
# HF_TOKEN = os.environ["HF_TOKEN"]
|
19 |
+
# headers = {"Authorization": f"Bearer {HF_TOKEN}"}
|
20 |
+
|
21 |
+
def translate(prompt_ , to_lang):
|
22 |
+
input_prompt = "translate this"
|
23 |
+
seed = 42
|
24 |
+
prompt = f"Instruction : Given an English input sentence translate it into {to_lang} sentence. \n input : \"{prompt_}\" \n {to_lang} : "
|
25 |
+
if len(prompt) == 0:
|
26 |
+
prompt = input_prompt
|
27 |
+
|
28 |
+
json_ = {
|
29 |
+
"inputs": prompt,
|
30 |
+
"parameters": {
|
31 |
+
"top_p": 0.9,
|
32 |
+
"temperature": 1.1,
|
33 |
+
"max_new_tokens": 250,
|
34 |
+
"return_full_text": False,
|
35 |
+
"do_sample": False,
|
36 |
+
"seed": seed,
|
37 |
+
"early_stopping": False,
|
38 |
+
"length_penalty": 0.0,
|
39 |
+
"eos_token_id": None,
|
40 |
+
},
|
41 |
+
"options": {
|
42 |
+
"use_cache": True,
|
43 |
+
"wait_for_model": True,
|
44 |
+
},
|
45 |
+
}
|
46 |
+
response = requests.request("POST", API_URL, json=json_) # headers=headers
|
47 |
+
# output = response.json()
|
48 |
+
output = json.loads(response.content.decode("utf-8"))
|
49 |
+
output_tmp = output[0]['generated_text']
|
50 |
+
solution = output_tmp.split(f"\n{to_lang}:")[0]
|
51 |
+
|
52 |
+
|
53 |
+
if '\n\n' in solution:
|
54 |
+
final_solution = solution.split("\n\n")[0]
|
55 |
+
else:
|
56 |
+
final_solution = solution
|
57 |
+
return final_solution
|
58 |
+
|
59 |
+
|
60 |
# Image to Text
|
61 |
def image_to_text(url):
|
62 |
# Load a transformer
|
|
|
145 |
except sqlite3.OperationalError as e:
|
146 |
st.error(f"An error occurred while trying to log in: {e}")
|
147 |
|
148 |
+
|
|
|
|
|
149 |
|
150 |
def predict(cap_col, target_language):
|
151 |
captions = []
|
|
|
153 |
pred_caption = image_to_text(url)
|
154 |
|
155 |
cap_col.markdown('#### Predicted Captions:')
|
156 |
+
translated_caption = translate(pred_caption, target_language)
|
157 |
captions.append(translated_caption)
|
158 |
|
159 |
for _ in range(4):
|
160 |
pred_caption = image_to_text(url)
|
161 |
if pred_caption not in captions:
|
162 |
+
translated_caption = translate(pred_caption, target_language)
|
163 |
captions.append(translated_caption)
|
164 |
|
165 |
cap_col.markdown('<div class="caption-container">', unsafe_allow_html=True)
|
|
|
178 |
img_upload = st.file_uploader(label='Upload Image', type=['jpg', 'png', 'jpeg'])
|
179 |
|
180 |
# Language selection dropdown
|
181 |
+
|
182 |
+
|
183 |
+
target_language = st.selectbox('Select Target Language', ['English', 'Spanish', 'Hindi', 'Italian'], index=0)
|
184 |
|
185 |
# Process image and generate captions
|
186 |
if img_url:
|