Spaces:
Runtime error
Runtime error
Create dataset.py
Browse files- dataset.py +56 -0
dataset.py
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from PIL import Image
|
3 |
+
import pandas as pd
|
4 |
+
|
5 |
+
|
6 |
+
|
7 |
+
class RetrievalDataset(torch.utils.data.Dataset):
|
8 |
+
def __init__(self, img_dir_path: str, annotations_file_path: str, split: str, transform=None, tokenizer=None) -> None:
|
9 |
+
self.img_dir_path = img_dir_path
|
10 |
+
self.transform = transform
|
11 |
+
self.tokenizer = tokenizer
|
12 |
+
self.split = split
|
13 |
+
self.annotations = self.split_data(
|
14 |
+
self.convert_image_names_to_path(
|
15 |
+
pd.read_csv(annotations_file_path)
|
16 |
+
)
|
17 |
+
)
|
18 |
+
|
19 |
+
def __len__(self) -> int:
|
20 |
+
return len(self.annotations)
|
21 |
+
|
22 |
+
def __getitem__(self, idx: int) -> tuple:
|
23 |
+
query_img_path = self.annotations.iloc[idx]['query_image']
|
24 |
+
query_text = self.annotations.iloc[idx]['query_text']
|
25 |
+
target_img_path = self.annotations.iloc[idx]['target_image']
|
26 |
+
query_img = Image.open(query_img_path).convert('RGB')
|
27 |
+
target_img = Image.open(target_img_path).convert('RGB')
|
28 |
+
# query_img = torchvision.io.read_image(path=query_img_path, mode=torchvision.io.image.ImageReadMode.RGB)
|
29 |
+
# target_img = torchvision.io.read_image(path=target_img_path, mode=torchvision.io.image.ImageReadMode.RGB)
|
30 |
+
if self.transform:
|
31 |
+
query_img = self.transform(query_img)
|
32 |
+
target_img = self.transform(target_img)
|
33 |
+
if self.tokenizer:
|
34 |
+
query_text = self.tokenizer(query_text).squeeze(0)
|
35 |
+
return query_img, query_text, target_img, self.annotations.iloc[idx]['query_text']
|
36 |
+
|
37 |
+
def split_data(self, annotations):
|
38 |
+
shuffled_df = annotations.sample(frac=1, random_state=42).reset_index(drop=True)
|
39 |
+
if self.split == "test":
|
40 |
+
return shuffled_df # sample test set
|
41 |
+
if self.split == "train":
|
42 |
+
return shuffled_df.iloc[:int(0.9 * len(shuffled_df))] # train set
|
43 |
+
if self.split == "validation":
|
44 |
+
return shuffled_df.iloc[int(0.9 * len(shuffled_df)):] # validation set
|
45 |
+
raise Exception("split is not valid")
|
46 |
+
|
47 |
+
def load_queries(self):
|
48 |
+
return self.annotations
|
49 |
+
|
50 |
+
def load_database(self):
|
51 |
+
return pd.DataFrame({'target_image': self.annotations["target_image"].unique()})
|
52 |
+
|
53 |
+
def convert_image_names_to_path(self, df):
|
54 |
+
df["query_image"] = self.img_dir_path + "/" + df["query_image"]
|
55 |
+
df["target_image"] = self.img_dir_path + "/" + df["target_image"]
|
56 |
+
return df
|