kusumakar commited on
Commit
8c70835
1 Parent(s): 57068ee

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +10 -11
app.py CHANGED
@@ -1,19 +1,19 @@
1
- import streamlit as st
2
  import numpy as np
3
  from PIL import Image
4
- from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel, GPT2Tokenizer, GPT2LMHeadModel
5
- import torch
6
- from transformers import BartTokenizer, BartForConditionalGeneration
7
-
8
- # Load pre-trained BART model and tokenizer
9
- tokenizer_2 = BartTokenizer.from_pretrained("facebook/bart-large-cnn")
10
- model_2 = BartForConditionalGeneration.from_pretrained("facebook/bart-large-cnn")
11
 
12
- # Directory path to the saved model on Google Drive
13
  model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
14
  feature_extractor = ViTFeatureExtractor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
15
  tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
16
 
 
 
 
 
 
17
  def generate_captions(image):
18
  image = Image.open(image).convert("RGB")
19
  generated_caption = tokenizer.decode(model.generate(feature_extractor(image, return_tensors="pt").pixel_values.to("cpu"))[0])
@@ -22,6 +22,7 @@ def generate_captions(image):
22
  generated_caption = sentence.replace(text_to_remove, "")
23
  return generated_caption
24
 
 
25
  def generate_paragraph(caption):
26
  # Tokenize the caption
27
  inputs = tokenizer_2([caption], max_length=1024, truncation=True, padding="longest", return_tensors="pt")
@@ -31,10 +32,8 @@ def generate_paragraph(caption):
31
 
32
  # Decode the generated output
33
  generated_text = tokenizer_2.decode(output[0], skip_special_tokens=True)
34
-
35
  return generated_text
36
 
37
-
38
  # create the Streamlit app
39
  def app():
40
  st.title('Image from your Side, Detailed description from my site')
 
1
+ import torch
2
  import numpy as np
3
  from PIL import Image
4
+ import streamlit as st
5
+ from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel, BartTokenizer, BartForConditionalGeneration
 
 
 
 
 
6
 
7
+ # pre-trained model to arrive at context
8
  model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
9
  feature_extractor = ViTFeatureExtractor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
10
  tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
11
 
12
+ # pre-trained to arrive at description
13
+ tokenizer_2 = BartTokenizer.from_pretrained("facebook/bart-large-cnn")
14
+ model_2 = BartForConditionalGeneration.from_pretrained("facebook/bart-large-cnn")
15
+
16
+ # function to generate context
17
  def generate_captions(image):
18
  image = Image.open(image).convert("RGB")
19
  generated_caption = tokenizer.decode(model.generate(feature_extractor(image, return_tensors="pt").pixel_values.to("cpu"))[0])
 
22
  generated_caption = sentence.replace(text_to_remove, "")
23
  return generated_caption
24
 
25
+ # function to generate description
26
  def generate_paragraph(caption):
27
  # Tokenize the caption
28
  inputs = tokenizer_2([caption], max_length=1024, truncation=True, padding="longest", return_tensors="pt")
 
32
 
33
  # Decode the generated output
34
  generated_text = tokenizer_2.decode(output[0], skip_special_tokens=True)
 
35
  return generated_text
36
 
 
37
  # create the Streamlit app
38
  def app():
39
  st.title('Image from your Side, Detailed description from my site')