Datasets:
Commit
·
1b771e8
1
Parent(s):
4ed1904
Update README.md
Browse files
README.md
CHANGED
@@ -155,62 +155,65 @@ In this respect, we carried out three actions:
|
|
155 |
## How to use
|
156 |
Data provided within this repository can be straightforwardly loaded via the *datasets* library as follows:
|
157 |
|
158 |
-
```
|
159 |
from datasets import load_dataset
|
160 |
dataset = load_dataset("MLNTeam-Unical/NFT-70M_transactions")
|
|
|
161 |
|
|
|
162 |
|
163 |
-
```
|
164 |
-
if you want to ....
|
165 |
-
```
|
166 |
from datasets import load_dataset
|
167 |
import numpy as np
|
168 |
|
169 |
|
170 |
transactions_dataset=load_dataset("MLNTeam-Unical/NFT-70M_transactions")
|
171 |
-
|
172 |
image_dataset=load_dataset("MLNTeam-Unical/NFT-70M_image")
|
173 |
-
|
174 |
text_dataset=load_dataset("MLNTeam-Unical/NFT-70M_text")
|
175 |
|
176 |
|
177 |
-
|
178 |
-
#compute a mapping from image_id to the row_index within the image dataset
|
179 |
image_id2row_index={int(id):k for k,id in enumerate(image_dataset["train"]["id"])}
|
180 |
-
|
|
|
181 |
text_id2row_index={int(id):k for k,id in enumerate(text_dataset["train"]["id"])}
|
182 |
|
183 |
|
184 |
def get_image_embedding(image_id,image_id2row_index,image_dataset):
|
185 |
-
#
|
186 |
idx_emb=image_id2row_index.get(int(image_id),None)
|
187 |
|
188 |
if idx_emb:
|
189 |
-
#
|
190 |
return np.array(image_dataset["train"].select([idx_emb])["emb"][0])
|
191 |
else:
|
192 |
-
#otherwise None is returned
|
193 |
return None
|
194 |
|
195 |
def get_text_embedding(text_id,text_id2row_index,text_dataset):
|
196 |
-
#
|
197 |
idx_emb=text_id2row_index.get(int(text_id),None)
|
|
|
198 |
if idx_emb:
|
199 |
-
#
|
200 |
return np.array(text_dataset["train"].select([idx_emb])["emb"][0])
|
201 |
else:
|
202 |
-
#otherwise None is returned
|
203 |
return None
|
204 |
-
#### USAGE EXAMPLE #########
|
205 |
|
|
|
|
|
|
|
206 |
transaction_id=120
|
207 |
|
|
|
208 |
id_image=transactions_dataset["train"].select([transaction_id])["collection_image"][0]
|
209 |
|
|
|
210 |
image_embedding=get_image_embedding(id_image,image_id2row_index,image_dataset)
|
211 |
|
|
|
212 |
id_text=transactions_dataset["train"].select([transaction_id])["collection_description"][0]
|
213 |
|
|
|
214 |
text_embedding=get_text_embedding(id_text,text_id2row_index,text_dataset)
|
215 |
```
|
216 |
|
|
|
155 |
## How to use
|
156 |
Data provided within this repository can be straightforwardly loaded via the *datasets* library as follows:
|
157 |
|
158 |
+
```python
|
159 |
from datasets import load_dataset
|
160 |
dataset = load_dataset("MLNTeam-Unical/NFT-70M_transactions")
|
161 |
+
```
|
162 |
|
163 |
+
Complementary data involving textual and visual embeddings can be integrated as follows:
|
164 |
|
165 |
+
```python
|
|
|
|
|
166 |
from datasets import load_dataset
|
167 |
import numpy as np
|
168 |
|
169 |
|
170 |
transactions_dataset=load_dataset("MLNTeam-Unical/NFT-70M_transactions")
|
|
|
171 |
image_dataset=load_dataset("MLNTeam-Unical/NFT-70M_image")
|
|
|
172 |
text_dataset=load_dataset("MLNTeam-Unical/NFT-70M_text")
|
173 |
|
174 |
|
175 |
+
# Mapping from image_id to the row_index within the image dataset
|
|
|
176 |
image_id2row_index={int(id):k for k,id in enumerate(image_dataset["train"]["id"])}
|
177 |
+
|
178 |
+
# Mapping from text_id to row_index within the text dataset
|
179 |
text_id2row_index={int(id):k for k,id in enumerate(text_dataset["train"]["id"])}
|
180 |
|
181 |
|
182 |
def get_image_embedding(image_id,image_id2row_index,image_dataset):
|
183 |
+
# If the mapping contains the image, the embedding exists
|
184 |
idx_emb=image_id2row_index.get(int(image_id),None)
|
185 |
|
186 |
if idx_emb:
|
187 |
+
# If the embedding exists, return it
|
188 |
return np.array(image_dataset["train"].select([idx_emb])["emb"][0])
|
189 |
else:
|
|
|
190 |
return None
|
191 |
|
192 |
def get_text_embedding(text_id,text_id2row_index,text_dataset):
|
193 |
+
# If the mapping contains the image, the embedding exists
|
194 |
idx_emb=text_id2row_index.get(int(text_id),None)
|
195 |
+
|
196 |
if idx_emb:
|
197 |
+
# If the embedding exists, return it
|
198 |
return np.array(text_dataset["train"].select([idx_emb])["emb"][0])
|
199 |
else:
|
|
|
200 |
return None
|
|
|
201 |
|
202 |
+
### USAGE EXAMPLE ###
|
203 |
+
|
204 |
+
# Select transaction_id
|
205 |
transaction_id=120
|
206 |
|
207 |
+
# Get the image_id (e.g., collection_image or nft_image)
|
208 |
id_image=transactions_dataset["train"].select([transaction_id])["collection_image"][0]
|
209 |
|
210 |
+
# Get the image
|
211 |
image_embedding=get_image_embedding(id_image,image_id2row_index,image_dataset)
|
212 |
|
213 |
+
# Get the text_id
|
214 |
id_text=transactions_dataset["train"].select([transaction_id])["collection_description"][0]
|
215 |
|
216 |
+
# Get the text
|
217 |
text_embedding=get_text_embedding(id_text,text_id2row_index,text_dataset)
|
218 |
```
|
219 |
|