Update app.py
Browse files
app.py
CHANGED
@@ -7,25 +7,8 @@ from PIL import Image
|
|
7 |
import torch.nn as nn
|
8 |
from transformers import Blip2Processor, Blip2ForConditionalGeneration, InstructBlipProcessor, InstructBlipForConditionalGeneration
|
9 |
from my_model.object_detection import detect_and_draw_objects
|
|
|
10 |
|
11 |
-
def load_caption_model(blip2=False, instructblip=True):
|
12 |
-
|
13 |
-
if blip2:
|
14 |
-
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b", load_in_8bit=True,torch_dtype=torch.float16)
|
15 |
-
model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", load_in_8bit=True,torch_dtype=torch.float16)
|
16 |
-
if torch.cuda.device_count() > 1:
|
17 |
-
model = nn.DataParallel(model)
|
18 |
-
model.to('cuda')
|
19 |
-
#model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16, device_map="auto")
|
20 |
-
|
21 |
-
if instructblip:
|
22 |
-
model = InstructBlipForConditionalGeneration.from_pretrained("Salesforce/instructblip-vicuna-7b", load_in_8bit=True,torch_dtype=torch.float16)
|
23 |
-
if torch.cuda.device_count() > 1:
|
24 |
-
model = nn.DataParallel(model)
|
25 |
-
model.to('cuda')
|
26 |
-
processor = InstructBlipProcessor.from_pretrained("Salesforce/instructblip-vicuna-7b", load_in_8bit=True,torch_dtype=torch.float16)
|
27 |
-
|
28 |
-
return model, processor
|
29 |
|
30 |
|
31 |
|
@@ -54,7 +37,16 @@ image = st.file_uploader("Upload an image", type=["png", "jpg", "jpeg"])
|
|
54 |
# Text input for the question
|
55 |
question = st.text_input("Enter your question about the image:")
|
56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
|
|
|
|
|
|
58 |
if st.button("Get Answer"):
|
59 |
if image is not None and question:
|
60 |
# Display the image
|
|
|
7 |
import torch.nn as nn
|
8 |
from transformers import Blip2Processor, Blip2ForConditionalGeneration, InstructBlipProcessor, InstructBlipForConditionalGeneration
|
9 |
from my_model.object_detection import detect_and_draw_objects
|
10 |
+
from my_model.captioner.image_captioning import get_caption
|
11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
|
14 |
|
|
|
37 |
# Text input for the question
|
38 |
question = st.text_input("Enter your question about the image:")
|
39 |
|
40 |
+
if st.button('Generate Caption'):
|
41 |
+
if image is not None:
|
42 |
+
# Display the image
|
43 |
+
st.image(image, use_column_width=True)
|
44 |
+
caption = get_caption(image)
|
45 |
+
st.write(caption)
|
46 |
|
47 |
+
else:
|
48 |
+
st.write("Please upload an image and enter a question.")
|
49 |
+
|
50 |
if st.button("Get Answer"):
|
51 |
if image is not None and question:
|
52 |
# Display the image
|