Update app.py
Browse files
app.py
CHANGED
@@ -1,56 +1,56 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
from model_architecture import ImageCaptionGenerationWithAttention
|
3 |
-
from transformers import BartForConditionalGeneration, BartTokenizer, ViTModel, ViTImageProcessor
|
4 |
-
import torch
|
5 |
-
from PIL import Image
|
6 |
-
from dotenv import load_dotenv
|
7 |
-
import os
|
8 |
-
import traceback
|
9 |
-
|
10 |
-
load_dotenv()
|
11 |
-
HF_TOKEN = os.getenv('hf_token')
|
12 |
-
|
13 |
-
|
14 |
-
class GenerateCaptions:
|
15 |
-
def __init__(self):
|
16 |
-
self.device = torch.device(
|
17 |
-
"cuda" if torch.cuda.is_available() else "cpu")
|
18 |
-
vit_model = ViTModel.from_pretrained(
|
19 |
-
"google/vit-base-patch16-224", token=HF_TOKEN).to(self.device)
|
20 |
-
bart_model = BartForConditionalGeneration.from_pretrained(
|
21 |
-
"facebook/bart-base").to(self.device)
|
22 |
-
self.processor = ViTImageProcessor.from_pretrained(
|
23 |
-
"google/vit-base-patch16-224")
|
24 |
-
self.tokenizer = BartTokenizer.from_pretrained("facebook/bart-base")
|
25 |
-
self.model = ImageCaptionGenerationWithAttention(
|
26 |
-
vit_model, bart_model, self.tokenizer)
|
27 |
-
self.model.load_state_dict(torch.load(
|
28 |
-
'image_captioning_model_state_dict.pt', map_location=self.device))
|
29 |
-
self.model.eval()
|
30 |
-
|
31 |
-
def generate_caption(self, frame, max_length=50, num_beams=5):
|
32 |
-
try:
|
33 |
-
image_pixel_values = self.processor(
|
34 |
-
frame, return_tensors="pt").pixel_values
|
35 |
-
generated_caption_ids = self.model.generate(
|
36 |
-
image_pixel_values, max_length, num_beams)
|
37 |
-
return self.tokenizer.decode(generated_caption_ids[0], skip_special_tokens=True)
|
38 |
-
except Exception as e:
|
39 |
-
print(e)
|
40 |
-
print(traceback.format_exc())
|
41 |
-
|
42 |
-
|
43 |
-
gc = GenerateCaptions()
|
44 |
-
|
45 |
-
demo = gr.Interface(
|
46 |
-
fn=gc.generate_caption,
|
47 |
-
inputs=gr.Image(type='pil'),
|
48 |
-
outputs="text",
|
49 |
-
title="Image Caption
|
50 |
-
examples=['Image.jpg', 'Image 2.jpg'],
|
51 |
-
submit_btn='Generate Caption',
|
52 |
-
flagging_mode='never'
|
53 |
-
)
|
54 |
-
|
55 |
-
|
56 |
-
demo.launch()
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from model_architecture import ImageCaptionGenerationWithAttention
|
3 |
+
from transformers import BartForConditionalGeneration, BartTokenizer, ViTModel, ViTImageProcessor
|
4 |
+
import torch
|
5 |
+
from PIL import Image
|
6 |
+
from dotenv import load_dotenv
|
7 |
+
import os
|
8 |
+
import traceback
|
9 |
+
|
10 |
+
load_dotenv()
|
11 |
+
HF_TOKEN = os.getenv('hf_token')
|
12 |
+
|
13 |
+
|
14 |
+
class GenerateCaptions:
|
15 |
+
def __init__(self):
|
16 |
+
self.device = torch.device(
|
17 |
+
"cuda" if torch.cuda.is_available() else "cpu")
|
18 |
+
vit_model = ViTModel.from_pretrained(
|
19 |
+
"google/vit-base-patch16-224", token=HF_TOKEN).to(self.device)
|
20 |
+
bart_model = BartForConditionalGeneration.from_pretrained(
|
21 |
+
"facebook/bart-base").to(self.device)
|
22 |
+
self.processor = ViTImageProcessor.from_pretrained(
|
23 |
+
"google/vit-base-patch16-224")
|
24 |
+
self.tokenizer = BartTokenizer.from_pretrained("facebook/bart-base")
|
25 |
+
self.model = ImageCaptionGenerationWithAttention(
|
26 |
+
vit_model, bart_model, self.tokenizer)
|
27 |
+
self.model.load_state_dict(torch.load(
|
28 |
+
'image_captioning_model_state_dict.pt', map_location=self.device))
|
29 |
+
self.model.eval()
|
30 |
+
|
31 |
+
def generate_caption(self, frame, max_length=50, num_beams=5):
|
32 |
+
try:
|
33 |
+
image_pixel_values = self.processor(
|
34 |
+
frame, return_tensors="pt").pixel_values
|
35 |
+
generated_caption_ids = self.model.generate(
|
36 |
+
image_pixel_values, max_length, num_beams)
|
37 |
+
return self.tokenizer.decode(generated_caption_ids[0], skip_special_tokens=True)
|
38 |
+
except Exception as e:
|
39 |
+
print(e)
|
40 |
+
print(traceback.format_exc())
|
41 |
+
|
42 |
+
|
43 |
+
gc = GenerateCaptions()
|
44 |
+
|
45 |
+
demo = gr.Interface(
|
46 |
+
fn=gc.generate_caption,
|
47 |
+
inputs=gr.Image(type='pil'),
|
48 |
+
outputs="text",
|
49 |
+
title="Image Caption Generation",
|
50 |
+
examples=['Image.jpg', 'Image 2.jpg'],
|
51 |
+
submit_btn='Generate Caption',
|
52 |
+
flagging_mode='never'
|
53 |
+
)
|
54 |
+
|
55 |
+
|
56 |
+
demo.launch()
|