Spaces:
Configuration error
Configuration error
app update
Browse files- .gitignore +3 -0
- README.md +3 -12
- app.py +53 -130
- downloader.py +66 -0
- test_inputs/bird_image.jpg +0 -0
- test_inputs/car.jpg +0 -3
- test_inputs/dog_audio.wav +0 -0
- test_inputs/dog_image.jpg +0 -0
- test_inputs/dragon.jpg +0 -0
.gitignore
CHANGED
@@ -33,3 +33,6 @@ gradio.egg-info/
|
|
33 |
|
34 |
# Virtual Environment
|
35 |
.env/
|
|
|
|
|
|
|
|
33 |
|
34 |
# Virtual Environment
|
35 |
.env/
|
36 |
+
|
37 |
+
#others
|
38 |
+
test_inputs/
|
README.md
CHANGED
@@ -1,13 +1,4 @@
|
|
1 |
-
|
2 |
-
|
3 |
-
|
4 |
-
colorFrom: indigo
|
5 |
-
colorTo: purple
|
6 |
-
sdk: gradio
|
7 |
-
sdk_version: 4.19.2
|
8 |
-
app_file: app.py
|
9 |
-
pinned: false
|
10 |
-
license: mit
|
11 |
-
---
|
12 |
|
13 |
-
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
1 |
+
# Imagebind demo
|
2 |
+
|
3 |
+
A gradio app showcasing multi-modal capabilities of Imagebind supported via lanceDB API
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
|
app.py
CHANGED
@@ -2,166 +2,89 @@ import lancedb
|
|
2 |
import lancedb.embeddings.imagebind
|
3 |
from lancedb.embeddings import get_registry
|
4 |
from lancedb.pydantic import LanceModel, Vector
|
5 |
-
import pandas as pd
|
6 |
-
import os
|
7 |
import gradio as gr
|
|
|
8 |
|
9 |
model = get_registry().get("imagebind").create()
|
10 |
|
|
|
11 |
class TextModel(LanceModel):
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
# "/test_inputs/bird_image.jpg"]
|
23 |
-
# audio_paths=["/test_inputs/dragon-growl-37570.wav",
|
24 |
-
# "/test_inputs/car_audio.wav",
|
25 |
-
# "/test_inputs/bird_audio.wav"]
|
26 |
-
|
27 |
-
image_paths=["./test_inputs/dragon.jpg",
|
28 |
-
"./test_inputs/car.jpg",
|
29 |
-
"./test_inputs/bird_image.jpg"]
|
30 |
-
audio_paths=["./test_inputs/dragon-growl-37570.wav",
|
31 |
-
"./test_inputs/car_audio.wav",
|
32 |
-
"./test_inputs/bird_audio.wav"]
|
33 |
# Load data
|
34 |
inputs = [
|
35 |
-
{
|
36 |
-
|
37 |
-
"audio_path":b,
|
38 |
-
"image_uri":c
|
39 |
-
} for a,
|
40 |
-
b,
|
41 |
-
c in zip(text_list,
|
42 |
-
audio_paths,
|
43 |
-
image_paths)
|
44 |
]
|
45 |
|
46 |
-
|
47 |
db = lancedb.connect("/tmp/lancedb")
|
48 |
-
table = db.create_table("img_bind",schema=TextModel)
|
49 |
table.add(inputs)
|
50 |
|
51 |
|
52 |
def process_image(inp_img) -> str:
|
53 |
-
|
54 |
actual = (
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
)
|
59 |
|
60 |
return actual.text, actual.audio_path
|
61 |
|
62 |
-
def process_text(inp_text) -> str:
|
63 |
|
|
|
64 |
actual = (
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
)
|
69 |
|
70 |
return actual.image_uri, actual.audio_path
|
71 |
|
72 |
-
def process_audio(inp_audio) -> str:
|
73 |
|
|
|
74 |
actual = (
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
)
|
79 |
|
80 |
return actual.image_uri, actual.text
|
81 |
|
82 |
-
css = """
|
83 |
-
output-audio, output-text {
|
84 |
-
display: none;
|
85 |
-
}
|
86 |
-
img {
|
87 |
-
width: 300px;
|
88 |
-
height: 450px;
|
89 |
-
object-fit: cover;
|
90 |
-
|
91 |
-
"""
|
92 |
-
with gr.Blocks(css=css) as app:
|
93 |
-
# Using Markdown for custom CSS (optional)
|
94 |
-
with gr.Tab("Image to Text and Audio"):
|
95 |
-
with gr.Row():
|
96 |
-
with gr.Column():
|
97 |
-
|
98 |
-
inp1 = gr.Image(value=image_paths[0],type='filepath',elem_id='img',interactive=False)
|
99 |
-
output_audio1 = gr.Audio(label="Output Audio", elem_id="output-audio")
|
100 |
-
output_text1 = gr.Textbox(label="Output Text", elem_id="output-text")
|
101 |
-
btn_img1 = gr.Button("Retrieve")
|
102 |
-
|
103 |
-
# output_audio1 = gr.Audio(label="Output Audio 1", elem_id="output-audio1")
|
104 |
-
with gr.Column():
|
105 |
-
inp2 = gr.Image(value=image_paths[1],type='filepath',elem_id='img',interactive=False)
|
106 |
-
output_audio2 = gr.Audio(label="Output Audio", elem_id="output-audio")
|
107 |
-
output_text2 = gr.Textbox(label="Output Text", elem_id="output-text")
|
108 |
-
btn_img2 = gr.Button("Retrieve")
|
109 |
-
|
110 |
-
with gr.Column():
|
111 |
-
inp3 = gr.Image(value=image_paths[2],type='filepath',elem_id='img',interactive=False)
|
112 |
-
output_audio3 = gr.Audio(label="Output Audio", elem_id="output-audio")
|
113 |
-
output_text3 = gr.Textbox(label="Output Text", elem_id="output-text")
|
114 |
-
btn_img3 = gr.Button("Retrieve")
|
115 |
-
|
116 |
-
with gr.Tab("Text to Image and Audio"):
|
117 |
-
with gr.Row():
|
118 |
-
with gr.Column():
|
119 |
-
input_txt1 = gr.Textbox(label="Enter a prompt:", elem_id="output-text")
|
120 |
-
output_audio4 = gr.Audio(label="Output Audio", elem_id="output-audio")
|
121 |
-
output_img1 = gr.Image(type='filepath',elem_id='img')
|
122 |
-
|
123 |
-
# with gr.Column():
|
124 |
-
# input_txt2 = gr.Textbox(label="Enter a prompt:", elem_id="output-text")
|
125 |
-
# output_audio5 = gr.Audio(label="Output Audio", elem_id="output-audio")
|
126 |
-
# output_img2 = gr.Image(type='filepath',elem_id='img')
|
127 |
-
|
128 |
-
|
129 |
-
# with gr.Column():
|
130 |
-
# input_txt3 = gr.Textbox(label="Enter a prompt:", elem_id="output-text")
|
131 |
-
# output_audio6 = gr.Audio(label="Output Audio", elem_id="output-audio")
|
132 |
-
# output_img3 = gr.Image(type='filepath',elem_id='img')
|
133 |
-
|
134 |
-
with gr.Tab("Audio to Image and Text"):
|
135 |
-
with gr.Row():
|
136 |
-
with gr.Column():
|
137 |
-
inp_audio1 = gr.Audio(value='./test_inputs/dragon-growl-37570.wav',type='filepath',elem_id='output-audio')
|
138 |
-
output_img7 = gr.Image(type='filepath',elem_id='img')
|
139 |
-
output_text7 = gr.Textbox(label="Output Text", elem_id="output-text")
|
140 |
-
btn_audio1 = gr.Button("Retrieve")
|
141 |
-
|
142 |
-
with gr.Column():
|
143 |
-
inp_audio2 = gr.Audio(value='./test_inputs/car_audio.wav',type='filepath',elem_id='output-audio')
|
144 |
-
output_img8 = gr.Image(type='filepath',elem_id='img')
|
145 |
-
output_text8 = gr.Textbox(label="Output Text", elem_id="output-text")
|
146 |
-
btn_audio2 = gr.Button("Retrieve")
|
147 |
-
|
148 |
-
with gr.Column():
|
149 |
-
inp_audio3 = gr.Audio(value='./test_inputs/bird_audio.wav',type='filepath',elem_id='output-audio')
|
150 |
-
output_img9 = gr.Image(type='filepath',elem_id='img')
|
151 |
-
output_text9 = gr.Textbox(label="Output Text", elem_id="output-text")
|
152 |
-
btn_audio3 = gr.Button("Retrieve")
|
153 |
-
|
154 |
-
# Click actions for buttons/Textboxes
|
155 |
-
btn_img1.click(process_image, inputs=[inp1],outputs=[output_text1,output_audio1])
|
156 |
-
btn_img2.click(process_image, inputs=[inp2],outputs=[output_text2,output_audio2])
|
157 |
-
btn_img3.click(process_image, inputs=[inp3],outputs=[output_text3,output_audio3])
|
158 |
-
|
159 |
-
input_txt1.submit(process_text, inputs=[input_txt1],outputs=[output_img1,output_audio4])
|
160 |
-
|
161 |
-
btn_audio1.click(process_audio, inputs=[inp_audio1],outputs=[output_img7,output_text7])
|
162 |
-
btn_audio2.click(process_audio, inputs=[inp_audio2],outputs=[output_img8,output_text8])
|
163 |
-
btn_audio3.click(process_audio, inputs=[inp_audio3],outputs=[output_img9,output_text9])
|
164 |
|
165 |
-
|
166 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
167 |
|
|
|
|
|
|
2 |
import lancedb.embeddings.imagebind
|
3 |
from lancedb.embeddings import get_registry
|
4 |
from lancedb.pydantic import LanceModel, Vector
|
|
|
|
|
5 |
import gradio as gr
|
6 |
+
from downloader import dowload_and_save_audio, dowload_and_save_image, base_path
|
7 |
|
8 |
model = get_registry().get("imagebind").create()
|
9 |
|
10 |
+
|
11 |
class TextModel(LanceModel):
|
12 |
+
text: str
|
13 |
+
image_uri: str = model.SourceField()
|
14 |
+
audio_path: str
|
15 |
+
vector: Vector(model.ndims()) = model.VectorField()
|
16 |
+
|
17 |
+
|
18 |
+
text_list = ["A bird", "A dragon", "A car"]
|
19 |
+
image_paths = dowload_and_save_image()
|
20 |
+
audio_paths = dowload_and_save_audio()
|
21 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
# Load data
|
23 |
inputs = [
|
24 |
+
{"text": a, "audio_path": b, "image_uri": c}
|
25 |
+
for a, b, c in zip(text_list, audio_paths, image_paths)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
]
|
27 |
|
|
|
28 |
db = lancedb.connect("/tmp/lancedb")
|
29 |
+
table = db.create_table("img_bind", schema=TextModel)
|
30 |
table.add(inputs)
|
31 |
|
32 |
|
33 |
def process_image(inp_img) -> str:
|
|
|
34 |
actual = (
|
35 |
+
table.search(inp_img, vector_column_name="vector")
|
36 |
+
.limit(1)
|
37 |
+
.to_pydantic(TextModel)[0]
|
38 |
)
|
39 |
|
40 |
return actual.text, actual.audio_path
|
41 |
|
|
|
42 |
|
43 |
+
def process_text(inp_text) -> str:
|
44 |
actual = (
|
45 |
+
table.search(inp_text, vector_column_name="vector")
|
46 |
+
.limit(1)
|
47 |
+
.to_pydantic(TextModel)[0]
|
48 |
)
|
49 |
|
50 |
return actual.image_uri, actual.audio_path
|
51 |
|
|
|
52 |
|
53 |
+
def process_audio(inp_audio) -> str:
|
54 |
actual = (
|
55 |
+
table.search(inp_audio, vector_column_name="vector")
|
56 |
+
.limit(1)
|
57 |
+
.to_pydantic(TextModel)[0]
|
58 |
)
|
59 |
|
60 |
return actual.image_uri, actual.text
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
|
63 |
+
im_to_at = gr.Interface(
|
64 |
+
process_image,
|
65 |
+
gr.Image(type="filepath", value=image_paths[0]),
|
66 |
+
[gr.Text(label="Output Text"), gr.Audio(label="Output Audio")],
|
67 |
+
examples=image_paths,
|
68 |
+
allow_flagging="never",
|
69 |
+
)
|
70 |
+
txt_to_ia = gr.Interface(
|
71 |
+
process_text,
|
72 |
+
gr.Textbox(label="Enter a prompt:"),
|
73 |
+
[gr.Image(label="Output Image"), gr.Audio(label="Output Audio")],
|
74 |
+
allow_flagging="never",
|
75 |
+
examples=text_list,
|
76 |
+
)
|
77 |
+
a_to_it = gr.Interface(
|
78 |
+
process_audio,
|
79 |
+
gr.Audio(type="filepath", value=audio_paths[0]),
|
80 |
+
[gr.Image(label="Output Image"), gr.Text(label="Output Text")],
|
81 |
+
examples=audio_paths,
|
82 |
+
allow_flagging="never",
|
83 |
+
)
|
84 |
+
demo = gr.TabbedInterface(
|
85 |
+
[im_to_at, txt_to_ia, a_to_it],
|
86 |
+
["Image to Text/Audio", "Text to Image/Audio", "Audio to Image/Text"],
|
87 |
+
)
|
88 |
|
89 |
+
if __name__ == "__main__":
|
90 |
+
demo.launch(share=True, allowed_paths=[f"{base_path}/test_inputs/"])
|
downloader.py
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import requests
|
2 |
+
import os
|
3 |
+
from pathlib import Path
|
4 |
+
|
5 |
+
# URL of the raw audio file on GitHub
|
6 |
+
audio_file_urls = [
|
7 |
+
"https://github.com/raghavdixit99/assets/raw/main/bird_audio.wav",
|
8 |
+
"https://github.com/raghavdixit99/assets/raw/main/dragon-growl-37570.wav",
|
9 |
+
"https://github.com/raghavdixit99/assets/raw/main/car_audio.wav",
|
10 |
+
]
|
11 |
+
image_urls = [
|
12 |
+
"https://github.com/raghavdixit99/assets/assets/34462078/abf47cc4-d979-4aaa-83be-53a2115bf318",
|
13 |
+
"https://github.com/raghavdixit99/assets/assets/34462078/93be928e-522b-4e37-889d-d4efd54b2112",
|
14 |
+
"https://github.com/raghavdixit99/assets/assets/34462078/025deaff-632a-4829-a86c-3de6e326402f",
|
15 |
+
]
|
16 |
+
|
17 |
+
base_path = os.path.dirname(os.path.abspath(__file__))
|
18 |
+
|
19 |
+
|
20 |
+
# Local path where you want to save the .wav file
|
21 |
+
def dowload_and_save_audio():
|
22 |
+
audio_pths = []
|
23 |
+
for url in audio_file_urls:
|
24 |
+
filename = url.split("/")[-1]
|
25 |
+
local_file_path = Path(f"{base_path}/test_inputs/{filename}")
|
26 |
+
local_file_path.parent.mkdir(parents=True, exist_ok=True)
|
27 |
+
# Perform the GET request
|
28 |
+
response = requests.get(url)
|
29 |
+
|
30 |
+
# Check if the request was successful
|
31 |
+
if response.status_code == 200:
|
32 |
+
# Write the content of the response to a local file
|
33 |
+
with open(local_file_path, "wb") as audio_file:
|
34 |
+
audio_file.write(response.content)
|
35 |
+
audio_pths.append(str(local_file_path))
|
36 |
+
print(
|
37 |
+
f"Audio file downloaded successfully and saved as '{local_file_path}'."
|
38 |
+
)
|
39 |
+
else:
|
40 |
+
print(f"Failed to download file. Status code: {response.status_code}")
|
41 |
+
return audio_pths
|
42 |
+
|
43 |
+
|
44 |
+
def dowload_and_save_image():
|
45 |
+
image_paths = []
|
46 |
+
for url in image_urls:
|
47 |
+
filename = url.split("/")[-1]
|
48 |
+
local_file_path = Path(f"{base_path}/test_inputs/{filename}.jpeg")
|
49 |
+
|
50 |
+
local_file_path.parent.mkdir(parents=True, exist_ok=True)
|
51 |
+
# Perform the GET request
|
52 |
+
response = requests.get(url)
|
53 |
+
|
54 |
+
# Check if the request was successful
|
55 |
+
if response.status_code == 200:
|
56 |
+
# Write the content of the response to a local file
|
57 |
+
with open(local_file_path, "wb") as image_file:
|
58 |
+
image_file.write(response.content)
|
59 |
+
image_paths.append(str(local_file_path))
|
60 |
+
print(
|
61 |
+
f"Image file downloaded successfully and saved as '{local_file_path}'."
|
62 |
+
)
|
63 |
+
else:
|
64 |
+
print(f"Failed to download file. Status code: {response.status_code}")
|
65 |
+
|
66 |
+
return image_paths
|
test_inputs/bird_image.jpg
DELETED
Binary file (21.4 kB)
|
|
test_inputs/car.jpg
DELETED
Git LFS Details
|
test_inputs/dog_audio.wav
DELETED
Binary file (461 kB)
|
|
test_inputs/dog_image.jpg
DELETED
Binary file (20.5 kB)
|
|
test_inputs/dragon.jpg
DELETED
Binary file (24.5 kB)
|
|