Spaces:
Build error
Build error
ampehta
commited on
Commit
โข
503acf7
1
Parent(s):
98b26e2
Revert "Merge branch 'main' of https://huggingface.co/spaces/flax-community/koclip into main"
Browse filesThis reverts commit 98b26e2c6b300d6deec2c1f5a119ad6089b11224, reversing
changes made to 699df87c9ee261cf6dfc69f6a6276d3e99bfbc3e.
- app.py +1 -0
- embed.py +11 -16
- image2text.py +2 -4
- text2image.py +8 -11
- utils.py +13 -11
app.py
CHANGED
@@ -3,6 +3,7 @@ import streamlit as st
|
|
3 |
import image2text
|
4 |
import text2image
|
5 |
|
|
|
6 |
PAGES = {"Text to Image": text2image, "Image to Text": image2text}
|
7 |
|
8 |
st.sidebar.title("Navigation")
|
|
|
3 |
import image2text
|
4 |
import text2image
|
5 |
|
6 |
+
|
7 |
PAGES = {"Text to Image": text2image, "Image to Text": image2text}
|
8 |
|
9 |
st.sidebar.title("Navigation")
|
embed.py
CHANGED
@@ -2,20 +2,21 @@ import argparse
|
|
2 |
import csv
|
3 |
import os
|
4 |
|
5 |
-
import jax.numpy as jnp
|
6 |
from PIL import Image
|
7 |
-
from tqdm import tqdm
|
8 |
|
9 |
from utils import load_model
|
|
|
|
|
|
|
|
|
10 |
|
11 |
|
12 |
def main(args):
|
13 |
root = args.image_path
|
14 |
files = list(os.listdir(root))
|
15 |
for f in files:
|
16 |
-
assert
|
17 |
for model_name in ["koclip-base", "koclip-large"]:
|
18 |
-
# for model_name in ["koclip-large"]:
|
19 |
model, processor = load_model(f"koclip/{model_name}")
|
20 |
with tqdm(total=len(files)) as pbar:
|
21 |
for counter in range(0, len(files), args.batch_size):
|
@@ -23,34 +24,28 @@ def main(args):
|
|
23 |
image_ids = []
|
24 |
for idx in range(counter, min(len(files), counter + args.batch_size)):
|
25 |
file_ = files[idx]
|
26 |
-
image = Image.open(os.path.join(root, file_)).convert(
|
27 |
images.append(image)
|
28 |
image_ids.append(file_)
|
29 |
|
30 |
pbar.update(args.batch_size)
|
31 |
try:
|
32 |
-
inputs = processor(
|
33 |
-
text=[""], images=images, return_tensors="jax", padding=True
|
34 |
-
)
|
35 |
except:
|
36 |
print(image_ids)
|
37 |
break
|
38 |
-
inputs[
|
39 |
-
inputs["pixel_values"], axes=[0, 2, 3, 1]
|
40 |
-
)
|
41 |
features = model(**inputs).image_embeds
|
42 |
with open(os.path.join(args.out_path, f"{model_name}.tsv"), "a+") as f:
|
43 |
writer = csv.writer(f, delimiter="\t")
|
44 |
for image_id, feature in zip(image_ids, features):
|
45 |
-
writer.writerow(
|
46 |
-
[image_id, ",".join(map(lambda x: str(x), feature))]
|
47 |
-
)
|
48 |
|
49 |
|
50 |
if __name__ == "__main__":
|
51 |
parser = argparse.ArgumentParser()
|
52 |
parser.add_argument("--batch_size", default=16)
|
53 |
-
parser.add_argument("--image_path", default="images
|
54 |
-
parser.add_argument("--out_path", default="features
|
55 |
args = parser.parse_args()
|
56 |
main(args)
|
|
|
2 |
import csv
|
3 |
import os
|
4 |
|
|
|
5 |
from PIL import Image
|
|
|
6 |
|
7 |
from utils import load_model
|
8 |
+
import jax.numpy as jnp
|
9 |
+
from jax import jit
|
10 |
+
|
11 |
+
from tqdm import tqdm
|
12 |
|
13 |
|
14 |
def main(args):
|
15 |
root = args.image_path
|
16 |
files = list(os.listdir(root))
|
17 |
for f in files:
|
18 |
+
assert(f[-4:] == ".jpg")
|
19 |
for model_name in ["koclip-base", "koclip-large"]:
|
|
|
20 |
model, processor = load_model(f"koclip/{model_name}")
|
21 |
with tqdm(total=len(files)) as pbar:
|
22 |
for counter in range(0, len(files), args.batch_size):
|
|
|
24 |
image_ids = []
|
25 |
for idx in range(counter, min(len(files), counter + args.batch_size)):
|
26 |
file_ = files[idx]
|
27 |
+
image = Image.open(os.path.join(root, file_)).convert('RGB')
|
28 |
images.append(image)
|
29 |
image_ids.append(file_)
|
30 |
|
31 |
pbar.update(args.batch_size)
|
32 |
try:
|
33 |
+
inputs = processor(text=[""], images=images, return_tensors="jax", padding=True)
|
|
|
|
|
34 |
except:
|
35 |
print(image_ids)
|
36 |
break
|
37 |
+
inputs['pixel_values'] = jnp.transpose(inputs['pixel_values'], axes=[0, 2, 3, 1])
|
|
|
|
|
38 |
features = model(**inputs).image_embeds
|
39 |
with open(os.path.join(args.out_path, f"{model_name}.tsv"), "a+") as f:
|
40 |
writer = csv.writer(f, delimiter="\t")
|
41 |
for image_id, feature in zip(image_ids, features):
|
42 |
+
writer.writerow([image_id, ",".join(map(lambda x: str(x), feature))])
|
|
|
|
|
43 |
|
44 |
|
45 |
if __name__ == "__main__":
|
46 |
parser = argparse.ArgumentParser()
|
47 |
parser.add_argument("--batch_size", default=16)
|
48 |
+
parser.add_argument("--image_path", default="images")
|
49 |
+
parser.add_argument("--out_path", default="features")
|
50 |
args = parser.parse_args()
|
51 |
main(args)
|
image2text.py
CHANGED
@@ -7,8 +7,6 @@ def app(model_name):
|
|
7 |
model, processor = load_model(model_name)
|
8 |
|
9 |
st.title("Image to Text")
|
10 |
-
st.markdown(
|
11 |
-
"""
|
12 |
Some text goes in here.
|
13 |
-
"""
|
14 |
-
)
|
|
|
7 |
model, processor = load_model(model_name)
|
8 |
|
9 |
st.title("Image to Text")
|
10 |
+
st.markdown("""
|
|
|
11 |
Some text goes in here.
|
12 |
+
""")
|
|
text2image.py
CHANGED
@@ -1,22 +1,21 @@
|
|
1 |
import os
|
2 |
|
3 |
-
import matplotlib.pyplot as plt
|
4 |
-
import numpy as np
|
5 |
import streamlit as st
|
6 |
|
7 |
-
from utils import
|
|
|
|
|
8 |
|
9 |
|
10 |
def app(model_name):
|
11 |
-
images_directory =
|
12 |
-
features_directory = f
|
13 |
|
14 |
files, index = load_index(features_directory)
|
15 |
-
model, processor = load_model(f
|
16 |
|
17 |
st.title("Text to Image Search Engine")
|
18 |
-
st.markdown(
|
19 |
-
"""
|
20 |
This demonstration explores capability of KoCLIP as a Korean-language Image search engine. Embeddings for each of
|
21 |
5000 images from [MSCOCO](https://cocodataset.org/#home) 2017 validation set was generated using trained KoCLIP
|
22 |
vision model. They are ranked based on cosine similarity distance from input Text query embeddings and top 10 images
|
@@ -28,11 +27,9 @@ def app(model_name):
|
|
28 |
Larger model `koclip-large` uses `klue/roberta` as text encoder and bigger `google/vit-large-patch16-224` as image encoder.
|
29 |
|
30 |
Example Queries : ์ํํธ(Apartment), ์๋์ฐจ(Car), ์ปดํจํฐ(Computer)
|
31 |
-
"""
|
32 |
-
)
|
33 |
|
34 |
query = st.text_input("ํ๊ธ ์ง๋ฌธ์ ์ ์ด์ฃผ์ธ์ (Korean Text Query) :", value="์ํํธ")
|
35 |
-
|
36 |
if st.button("์ง๋ฌธ (Query)"):
|
37 |
proc = processor(text=[query], images=None, return_tensors="jax", padding=True)
|
38 |
vec = np.asarray(model.get_text_features(**proc))
|
|
|
1 |
import os
|
2 |
|
|
|
|
|
3 |
import streamlit as st
|
4 |
|
5 |
+
from utils import load_model, load_index
|
6 |
+
import numpy as np
|
7 |
+
import matplotlib.pyplot as plt
|
8 |
|
9 |
|
10 |
def app(model_name):
|
11 |
+
images_directory = 'images/val2017'
|
12 |
+
features_directory = f'features/val2017/{model_name}.tsv'
|
13 |
|
14 |
files, index = load_index(features_directory)
|
15 |
+
model, processor = load_model(f'koclip/{model_name}')
|
16 |
|
17 |
st.title("Text to Image Search Engine")
|
18 |
+
st.markdown("""
|
|
|
19 |
This demonstration explores capability of KoCLIP as a Korean-language Image search engine. Embeddings for each of
|
20 |
5000 images from [MSCOCO](https://cocodataset.org/#home) 2017 validation set was generated using trained KoCLIP
|
21 |
vision model. They are ranked based on cosine similarity distance from input Text query embeddings and top 10 images
|
|
|
27 |
Larger model `koclip-large` uses `klue/roberta` as text encoder and bigger `google/vit-large-patch16-224` as image encoder.
|
28 |
|
29 |
Example Queries : ์ํํธ(Apartment), ์๋์ฐจ(Car), ์ปดํจํฐ(Computer)
|
30 |
+
""")
|
|
|
31 |
|
32 |
query = st.text_input("ํ๊ธ ์ง๋ฌธ์ ์ ์ด์ฃผ์ธ์ (Korean Text Query) :", value="์ํํธ")
|
|
|
33 |
if st.button("์ง๋ฌธ (Query)"):
|
34 |
proc = processor(text=[query], images=None, return_tensors="jax", padding=True)
|
35 |
vec = np.asarray(model.get_text_features(**proc))
|
utils.py
CHANGED
@@ -1,28 +1,26 @@
|
|
1 |
import nmslib
|
2 |
-
import numpy as np
|
3 |
import streamlit as st
|
4 |
-
from transformers import
|
|
|
5 |
|
6 |
from koclip import FlaxHybridCLIP
|
7 |
|
8 |
-
|
9 |
@st.cache(allow_output_mutation=True)
|
10 |
def load_index(img_file):
|
11 |
filenames, embeddings = [], []
|
12 |
lines = open(img_file, "r")
|
13 |
for line in lines:
|
14 |
-
cols = line.strip().split(
|
15 |
filename = cols[0]
|
16 |
-
embedding = [float(x) for x in cols[1].split(
|
17 |
filenames.append(filename)
|
18 |
embeddings.append(embedding)
|
19 |
embeddings = np.array(embeddings)
|
20 |
-
index = nmslib.init(method=
|
21 |
index.addDataPointBatch(embeddings)
|
22 |
-
index.createIndex({
|
23 |
return filenames, index
|
24 |
|
25 |
-
|
26 |
@st.cache(allow_output_mutation=True)
|
27 |
def load_model(model_name="koclip/koclip-base"):
|
28 |
assert model_name in {"koclip/koclip-base", "koclip/koclip-large"}
|
@@ -30,7 +28,11 @@ def load_model(model_name="koclip/koclip-base"):
|
|
30 |
processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
|
31 |
processor.tokenizer = AutoTokenizer.from_pretrained("klue/roberta-large")
|
32 |
if model_name == "koclip/koclip-large":
|
33 |
-
processor.feature_extractor = ViTFeatureExtractor.from_pretrained(
|
34 |
-
|
35 |
-
|
|
|
|
|
|
|
|
|
36 |
return model, processor
|
|
|
1 |
import nmslib
|
|
|
2 |
import streamlit as st
|
3 |
+
from transformers import CLIPProcessor, AutoTokenizer, ViTFeatureExtractor
|
4 |
+
import numpy as np
|
5 |
|
6 |
from koclip import FlaxHybridCLIP
|
7 |
|
|
|
8 |
@st.cache(allow_output_mutation=True)
|
9 |
def load_index(img_file):
|
10 |
filenames, embeddings = [], []
|
11 |
lines = open(img_file, "r")
|
12 |
for line in lines:
|
13 |
+
cols = line.strip().split('\t')
|
14 |
filename = cols[0]
|
15 |
+
embedding = np.array([float(x) for x in cols[1].split(',')])
|
16 |
filenames.append(filename)
|
17 |
embeddings.append(embedding)
|
18 |
embeddings = np.array(embeddings)
|
19 |
+
index = nmslib.init(method='hnsw', space='cosinesimil')
|
20 |
index.addDataPointBatch(embeddings)
|
21 |
+
index.createIndex({'post': 2}, print_progress=True)
|
22 |
return filenames, index
|
23 |
|
|
|
24 |
@st.cache(allow_output_mutation=True)
|
25 |
def load_model(model_name="koclip/koclip-base"):
|
26 |
assert model_name in {"koclip/koclip-base", "koclip/koclip-large"}
|
|
|
28 |
processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
|
29 |
processor.tokenizer = AutoTokenizer.from_pretrained("klue/roberta-large")
|
30 |
if model_name == "koclip/koclip-large":
|
31 |
+
processor.feature_extractor = ViTFeatureExtractor.from_pretrained("google/vit-large-patch16-224")
|
32 |
+
return model, processor
|
33 |
+
|
34 |
+
@st.cache(allow_output_mutation=True)
|
35 |
+
def load_model_v2(model_name="koclip/koclip"):
|
36 |
+
model = FlaxHybridCLIP.from_pretrained(model_name)
|
37 |
+
processor = CLIPProcessor.from_pretrained(model_name)
|
38 |
return model, processor
|