Update dataset card for ReT-M2KR: Add task category, links, and sample usage

README.md

@@ -1,16 +1,31 @@
 ---
 license: mit
+task_categories:
+- visual-document-retrieval
+language:
+- en
+tags:
+- multimodal
+- retrieval
+- image-text-retrieval
+- rag
+- vqa
+- m2kr
+- m-beir
+- vision-language
 ---
 
-The dataset used to train and evaluate [ReT](https://www.arxiv.org/abs/2503.01980) for multimodal information retrieval. The dataset is almost the same as the original [M2KR](https://huggingface.co/datasets/BByrneLab/multi_task_multi_modal_knowledge_retrieval_benchmark_M2KR), with a few modifications:
-- we exlude any data from MSMARCO, as it does not contain query images;
-- we add passage images to OVEN, InfoSeek, E-VQA, and OKVQA. Refer to the paper for more details.
+# ReT-M2KR Dataset
 
+This repository contains the dataset used to train and evaluate **ReT-2: Recurrence Meets Transformers for Universal Multimodal Retrieval**. ReT-2 is a unified retrieval model designed to support multimodal queries (composed of images and text) and search across multimodal document collections where text and images coexist.
 
-## Sources
-- **Repository:** https://github.com/aimagelab/ReT
-- **Paper:** [Recurrence-Enhanced Vision-and-Language Transformers for Robust Multimodal Document Retrieval](https://www.arxiv.org/abs/2503.01980) (CVPR 2025)
+Paper: [Recurrence Meets Transformers for Universal Multimodal Retrieval](https://huggingface.co/papers/2509.08897)
+Code: https://github.com/aimagelab/ReT-2
 
+This dataset is a modified version of the original [M2KR](https://huggingface.co/datasets/BByrneLab/multi_task_multi_modal_knowledge_retrieval_benchmark_M2KR) benchmark, specifically adapted for ReT-2. The modifications include:
+-   Excluding any data from MSMARCO, as it does not contain query images.
+-   Adding passage images to OVEN, InfoSeek, E-VQA, and OKVQA.
+Refer to the paper for more details.
 
 ## Download images
 1. Initialize git LFS
@@ -35,20 +50,74 @@ cd ../images/evqa_kb
 cat evqa-kb-img-{00000..00241}.tar.gz | tar xzf -
 ```
 
+## Sample Usage
 
-## RAG - InfoSeek
-`jsonl/rag/kb_infoseek525k.jsonl` is the knowledge base used to execute experiments on Retrieval-Augmented Generation on the InfoSeek benchmark. The field `passage_image_path` contains a relative path to the Wikipedia image  associated with a given passage. The Wikipedia images can be downloaded from the [OVEN](https://huggingface.co/datasets/ychenNLP/oven/blob/main/all_wikipedia_images.tar) repository.  
+Here's an example of how to use ReT-2 with 🤗's Transformers to compute query-passage similarity:
 
-## Citation
+```python
+from src.models import Ret2Model
+import requests
+from PIL import Image
+from io import BytesIO
+import torch
+import torch.nn.functional as F
+
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+
+headers = {
+    'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
+}
+
+query_img_url = 'https://upload.wikimedia.org/wikipedia/commons/8/84/Ghirlandina_%28Modena%29.jpg'
+response = requests.get(query_img_url, headers=headers)
+query_image = Image.open(BytesIO(response.content)).convert('RGB')
+query_text = 'Where is this building located?'
+
+passage_img_url = 'https://upload.wikimedia.org/wikipedia/commons/0/09/Absidi_e_Ghirlandina.jpg'
+response = requests.get(query_img_url, headers=headers)
+passage_image = Image.open(BytesIO(response.content)).convert('RGB')
+passage_text = (
+    "The Ghirlandina is the bell tower of the Cathedral of Modena, in Modena, Italy. "
+    "It is 86.12 metres (282.7 ft) high and is the symbol of the city. "
+    "It was built in Romanesque style in the 12th century and is part of a UNESCO World Heritage Site."
+)
+
+model = Ret2Model.from_pretrained('aimagelab/ReT2-M2KR-ColBERT-SigLIP2-ViT-L', device_map=device)
 
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+query_txt_inputs = model.tokenizer([query_text], return_tensors='pt').to(device)
+query_img_inputs = model.image_processor([query_image], return_tensors='pt').to(device)
+passage_txt_inputs = model.tokenizer([passage_text], return_tensors='pt').to(device)
+passage_img_inputs = model.image_processor([passage_image], return_tensors='pt').to(device)
 
-**BibTeX:**
+with torch.inference_mode():
+    query_feats = model.get_ret_features(
+        input_ids=query_txt_inputs.input_ids,
+        attention_mask=query_txt_inputs.attention_mask,
+        pixel_values=query_img_inputs.pixel_values
+    )
+
+    passage_feats = model.get_ret_features(
+        input_ids=passage_txt_inputs.input_ids,
+        attention_mask=passage_txt_inputs.attention_mask,
+        pixel_values=passage_img_inputs.pixel_values
+    )
+
+    sim = F.normalize(query_feats, p=2, dim=-1) @ F.normalize(passage_feats, p=2, dim=-1).T
+
+print(f"query-passage similarity: {sim.item():.3f}")
 ```
-@inproceedings{caffagni2025recurrence,
-  title={{Recurrence-Enhanced Vision-and-Language Transformers for Robust Multimodal Document Retrieval}},
-  author={Caffagni, Davide and Sarto, Sara and Cornia, Marcella and Baraldi, Lorenzo and Cucchiara, Rita},
-  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
-  year={2025}
+
+## RAG - InfoSeek
+`jsonl/rag/kb_infoseek525k.jsonl` is the knowledge base used to execute experiments on Retrieval-Augmented Generation on the InfoSeek benchmark. The field `passage_image_path` contains a relative path to the Wikipedia image associated with a given passage. The Wikipedia images can be downloaded from the [OVEN](https://huggingface.co/datasets/ychenNLP/oven/blob/main/all_wikipedia_images.tar) repository.
+
+## Citation
+
+If you happen to use our work, please cite it with the following BibTeX:
+```bibtex
+@article{caffagni2025recurrencemeetstransformers,
+      title={{Recurrence Meets Transformers for Universal Multimodal Retrieval}},
+      author={Davide Caffagni and Sara Sarto and Marcella Cornia and Lorenzo Baraldi and Rita Cucchiara},
+      journal={arXiv preprint arXiv:2509.08897},
+      year={2025}
 }
 ```