Update README.md
Browse files
README.md
CHANGED
|
@@ -18,8 +18,12 @@ base_model_relation: finetune
|
|
| 18 |
<img width="500" alt="image" src="https://cdn-uploads.huggingface.co/production/uploads/64006c09330a45b03605bba3/zJsd2hqd3EevgXo6fNgC-.png">
|
| 19 |
</div>
|
| 20 |
|
|
|
|
|
|
|
| 21 |
# Introduction
|
| 22 |
|
|
|
|
|
|
|
| 23 |
We are excited to announce the release of **`TokenOCR`**, the first token-level visual foundation model specifically tailored for text-image-related tasks,
|
| 24 |
designed to support a variety of traditional downstream applications. To facilitate the pretraining of TokenOCR,
|
| 25 |
we also devise a high-quality data production pipeline that constructs the first token-level image text dataset,
|
|
@@ -27,8 +31,12 @@ we also devise a high-quality data production pipeline that constructs the first
|
|
| 27 |
Furthermore, leveraging this foundation with exceptional image-as-text capability,
|
| 28 |
we seamlessly replace previous VFMs with TokenOCR to construct a document-level MLLM, **`TokenVL`**, for VQA-based document understanding tasks.
|
| 29 |
|
|
|
|
|
|
|
| 30 |
# Token Family
|
| 31 |
|
|
|
|
|
|
|
| 32 |
<!-- ## TokenIT -->
|
| 33 |
<h2 style="color: #4CAF50;">TokenIT</h2>
|
| 34 |
|
|
@@ -146,9 +154,12 @@ Please refer to our technical report for more details.
|
|
| 146 |
|
| 147 |
<!-- 
|
| 148 |
-->
|
|
|
|
|
|
|
| 149 |
<!-- ## TokenVL -->
|
| 150 |
<h2 style="color: #4CAF50;">TokenVL</h2>
|
| 151 |
|
|
|
|
| 152 |
we employ the TokenOCR as the visual foundation model and further develop an MLLM, named TokenVL, tailored for document understanding.
|
| 153 |
Following the previous training paradigm, TokenVL also includes two stages:
|
| 154 |
|
|
|
|
| 18 |
<img width="500" alt="image" src="https://cdn-uploads.huggingface.co/production/uploads/64006c09330a45b03605bba3/zJsd2hqd3EevgXo6fNgC-.png">
|
| 19 |
</div>
|
| 20 |
|
| 21 |
+
<center>
|
| 22 |
+
|
| 23 |
# Introduction
|
| 24 |
|
| 25 |
+
</center>
|
| 26 |
+
|
| 27 |
We are excited to announce the release of **`TokenOCR`**, the first token-level visual foundation model specifically tailored for text-image-related tasks,
|
| 28 |
designed to support a variety of traditional downstream applications. To facilitate the pretraining of TokenOCR,
|
| 29 |
we also devise a high-quality data production pipeline that constructs the first token-level image text dataset,
|
|
|
|
| 31 |
Furthermore, leveraging this foundation with exceptional image-as-text capability,
|
| 32 |
we seamlessly replace previous VFMs with TokenOCR to construct a document-level MLLM, **`TokenVL`**, for VQA-based document understanding tasks.
|
| 33 |
|
| 34 |
+
<center>
|
| 35 |
+
|
| 36 |
# Token Family
|
| 37 |
|
| 38 |
+
</center>
|
| 39 |
+
|
| 40 |
<!-- ## TokenIT -->
|
| 41 |
<h2 style="color: #4CAF50;">TokenIT</h2>
|
| 42 |
|
|
|
|
| 154 |
|
| 155 |
<!-- 
|
| 156 |
-->
|
| 157 |
+
|
| 158 |
+
<center>
|
| 159 |
<!-- ## TokenVL -->
|
| 160 |
<h2 style="color: #4CAF50;">TokenVL</h2>
|
| 161 |
|
| 162 |
+
</center>
|
| 163 |
we employ the TokenOCR as the visual foundation model and further develop an MLLM, named TokenVL, tailored for document understanding.
|
| 164 |
Following the previous training paradigm, TokenVL also includes two stages:
|
| 165 |
|