deploy: 950fdb5

RapidAI · Nov 30, 2023 · 55d5093 · 55d5093
1 parent 422a1ae
commit 55d5093
Show file tree

Hide file tree

Showing 18 changed files with 1,443 additions and 56 deletions.
diff --git a/docs/blog/index.html b/docs/blog/index.html
@@ -156,6 +156,11 @@
                                             <li class=" "><a class="sidebar-nested-link" href="https://rapidai.github.io/TableStructureRec/docs/blog/wired_table_rec/">Cycle-CenterNet: 有线表格结构识别算法</a></li>
 
 
+
+
+                                            <li class=" "><a class="sidebar-nested-link" href="https://rapidai.github.io/TableStructureRec/docs/blog/table_rec_evaluate/">三个表格识别算法评测</a></li>
+
+
                                 </ul>
                             </div>
                         </li>
@@ -410,6 +415,23 @@ <h1 class="content-title mb-0">
             </a>
         </div>
 
+        <div id="list-item" class="col-md-4 col-12 mt-4 pt-2">
+            <a class="text-decoration-none text-reset" href="https://rapidai.github.io/TableStructureRec/docs/blog/table_rec_evaluate/">
+                <div class="card h-100 features feature-full-bg rounded p-4 position-relative overflow-hidden border-1">
+                    <span class="icon-color d-flex my-3">
+                        <i class="material-icons align-middle">table</i>
+
+
+                    </span>
+                    <div class="card-body p-0 content">
+                        <p class="fs-5 fw-semibold card-title mb-1">三个表格识别算法评测</p>
+                        <p class="para card-text mb-0"></p>
+                    </div>
+
+                </div>
+            </a>
+        </div>
+
     </div>
 
                                             </div>
@@ -651,6 +673,13 @@ <h1 class="content-title mb-0">
 
 
 
+
+
+
+
+
+
+
 
 
 
@@ -740,6 +769,15 @@ <h1 class="content-title mb-0">
     index.add(
             {
                 id:  8 ,
+                href: "\/TableStructureRec\/docs\/blog\/table_rec_evaluate\/",
+                title: "三个表格识别算法评测",
+                description: "引言 link为了便于比较不同表格识别算法的效果差异，本篇文章基于评测工具TableRecognitionMetric和表格测试数据集liekkas/table_recognition上计算不同算法的TEDS指标。\n指标结果 link 方法 TEDS RapidTable 0.58786 lineless_table_rec 0.50054 wired_table_rec 0.63316 评测步骤 link1. 安装评测数据集和评测工具包 link pip install table_recognition_metric pip install modelscope==1.5.2 pip install rapidocr_onnxruntime==1.3.8 2. 安装表格识别推理库 link pip install rapid_table pip install lineless_table_rec pip install wired_table_rec 3. 推理代码 link info 完整评测代码，请移步Gist\nfrom modelscope.msdatasets import MsDataset from rapid_table import RapidTable from lineless_table_rec import LinelessTableRecognition from wired_table_rec import WiredTableRecognition from table_recognition_metric import TEDS test_data = MsDataset.load( \"table_recognition\", namespace=\"liekkas\", subset_name=\"default\", split=\"test\", ) # 这里依次更换不同算法实例即可 table_engine = RapidTable() # table_engine = LinelessTableRecognition() # table_engine = WiredTableRecognition() teds = TEDS() content = [] for one_data in test_data: img_path = one_data.",
+                content: "引言 link为了便于比较不同表格识别算法的效果差异，本篇文章基于评测工具TableRecognitionMetric和表格测试数据集liekkas/table_recognition上计算不同算法的TEDS指标。\n指标结果 link 方法 TEDS RapidTable 0.58786 lineless_table_rec 0.50054 wired_table_rec 0.63316 评测步骤 link1. 安装评测数据集和评测工具包 link pip install table_recognition_metric pip install modelscope==1.5.2 pip install rapidocr_onnxruntime==1.3.8 2. 安装表格识别推理库 link pip install rapid_table pip install lineless_table_rec pip install wired_table_rec 3. 推理代码 link info 完整评测代码，请移步Gist\nfrom modelscope.msdatasets import MsDataset from rapid_table import RapidTable from lineless_table_rec import LinelessTableRecognition from wired_table_rec import WiredTableRecognition from table_recognition_metric import TEDS test_data = MsDataset.load( \"table_recognition\", namespace=\"liekkas\", subset_name=\"default\", split=\"test\", ) # 这里依次更换不同算法实例即可 table_engine = RapidTable() # table_engine = LinelessTableRecognition() # table_engine = WiredTableRecognition() teds = TEDS() content = [] for one_data in test_data: img_path = one_data.get(\"image:FILE\") gt = one_data.get(\"label\") pred_str, _ = table_engine(img_path) scores = teds(gt, pred_str) content.append(scores) print(f\"{img_path}\\t{scores:.5f}\") avg = sum(content) / len(content) print(f'{avg:.5f}') 4. 写在最后 link以上评测仅是基于表格测试数据集liekkas/table_recognition测试而来，不能完全代表模型效果。\n因为每个模型训练数据不同，测试数据集如与训练数据相差较大，难免效果较差，请针对自身场景客观看待评测指标。\n"
+            }
+        );
+    index.add(
+            {
+                id:  9 ,
                 href: "\/TableStructureRec\/docs\/sponsor\/",
                 title: "给作者加油",
                 description: "写在前面 linkI like open source and AI technology because I think open source and AI will bring convenience and help to people in need, and will also make the world a better place. By donating to these projects, you can join me in making AI bring warmth and beauty to more people.\n我喜欢开源，喜欢AI技术，因为我认为开源和AI会为有需要的人带来方便和帮助，也会让这个世界变得更好。通过对这些项目的捐赠，您可以和我一道让AI为更多人带来温暖和美好。\n知识星球RapidAI私享群 link这里的提问会优先得到回答和支持，也会享受到RapidAI组织后续持续优质的服务，欢迎大家的加入。\n支付宝或微信打赏 (Alipay reward or WeChat reward) link通过支付宝或者微信给作者打赏，请写好备注。 Give the author a reward through Alipay or WeChat.",
@@ -748,7 +786,7 @@ <h1 class="content-title mb-0">
         );
     index.add(
             {
-                id:  9 ,
+                id:  10 ,
                 href: "\/TableStructureRec\/docs\/faq\/",
                 title: "常见问题",
                 description: "",
@@ -757,7 +795,7 @@ <h1 class="content-title mb-0">
         );
     index.add(
             {
-                id:  10 ,
+                id:  11 ,
                 href: "\/TableStructureRec\/docs\/changelog\/",
                 title: "更新日志",
                 description: "2023-11-22 update: link 发布wired_table_rec==0.0.1包 2023-11-20 update: link 发布lineless_table_rec==0.0.1包 2023-11-19 update: link 添加无线表格识别算法（LORE）的推理代码 ",
@@ -766,7 +804,7 @@ <h1 class="content-title mb-0">
         );
     index.add(
             {
-                id:  11 ,
+                id:  12 ,
                 href: "\/TableStructureRec\/docs\/",
                 title: "Docs",
                 description: "",

diff --git a/docs/blog/index.xml b/docs/blog/index.xml
@@ -37,5 +37,16 @@
 参考资料 link 读光-表格结构识别-有线表格 </description>
     </item>
 
+    <item>
+      <title>三个表格识别算法评测</title>
+      <link>https://rapidai.github.io/TableStructureRec/docs/blog/table_rec_evaluate/</link>
+      <pubDate>Thu, 30 Nov 2023 00:00:00 +0000</pubDate>
+
+      <guid>https://rapidai.github.io/TableStructureRec/docs/blog/table_rec_evaluate/</guid>
+      <description>引言 link为了便于比较不同表格识别算法的效果差异，本篇文章基于评测工具TableRecognitionMetric和表格测试数据集liekkas/table_recognition上计算不同算法的TEDS指标。
+指标结果 link 方法 TEDS RapidTable 0.58786 lineless_table_rec 0.50054 wired_table_rec 0.63316 评测步骤 link1. 安装评测数据集和评测工具包 link pip install table_recognition_metric pip install modelscope==1.5.2 pip install rapidocr_onnxruntime==1.3.8 2. 安装表格识别推理库 link pip install rapid_table pip install lineless_table_rec pip install wired_table_rec 3. 推理代码 link info 完整评测代码，请移步Gist
+from modelscope.msdatasets import MsDataset from rapid_table import RapidTable from lineless_table_rec import LinelessTableRecognition from wired_table_rec import WiredTableRecognition from table_recognition_metric import TEDS test_data = MsDataset.load( &amp;#34;table_recognition&amp;#34;, namespace=&amp;#34;liekkas&amp;#34;, subset_name=&amp;#34;default&amp;#34;, split=&amp;#34;test&amp;#34;, ) # 这里依次更换不同算法实例即可 table_engine = RapidTable() # table_engine = LinelessTableRecognition() # table_engine = WiredTableRecognition() teds = TEDS() content = [] for one_data in test_data: img_path = one_data.</description>
+    </item>
+
   </channel>
 </rss>
diff --git a/docs/blog/lineless_table_rec/index.html b/docs/blog/lineless_table_rec/index.html
@@ -170,6 +170,11 @@
                                             <li class=" "><a class="sidebar-nested-link" href="https://rapidai.github.io/TableStructureRec/docs/blog/wired_table_rec/">Cycle-CenterNet: 有线表格结构识别算法</a></li>
 
 
+
+
+                                            <li class=" "><a class="sidebar-nested-link" href="https://rapidai.github.io/TableStructureRec/docs/blog/table_rec_evaluate/">三个表格识别算法评测</a></li>
+
+
                                 </ul>
                             </div>
                         </li>
@@ -751,6 +756,13 @@ <h3 id="参考资料">参考资料 <a href="#%e5%8f%82%e8%80%83%e8%b5%84%e6%96%9
 
 
 
+
+
+
+
+
+
+
 
 
 
@@ -840,6 +852,15 @@ <h3 id="参考资料">参考资料 <a href="#%e5%8f%82%e8%80%83%e8%b5%84%e6%96%9
     index.add(
             {
                 id:  8 ,
+                href: "\/TableStructureRec\/docs\/blog\/table_rec_evaluate\/",
+                title: "三个表格识别算法评测",
+                description: "引言 link为了便于比较不同表格识别算法的效果差异，本篇文章基于评测工具TableRecognitionMetric和表格测试数据集liekkas/table_recognition上计算不同算法的TEDS指标。\n指标结果 link 方法 TEDS RapidTable 0.58786 lineless_table_rec 0.50054 wired_table_rec 0.63316 评测步骤 link1. 安装评测数据集和评测工具包 link pip install table_recognition_metric pip install modelscope==1.5.2 pip install rapidocr_onnxruntime==1.3.8 2. 安装表格识别推理库 link pip install rapid_table pip install lineless_table_rec pip install wired_table_rec 3. 推理代码 link info 完整评测代码，请移步Gist\nfrom modelscope.msdatasets import MsDataset from rapid_table import RapidTable from lineless_table_rec import LinelessTableRecognition from wired_table_rec import WiredTableRecognition from table_recognition_metric import TEDS test_data = MsDataset.load( \"table_recognition\", namespace=\"liekkas\", subset_name=\"default\", split=\"test\", ) # 这里依次更换不同算法实例即可 table_engine = RapidTable() # table_engine = LinelessTableRecognition() # table_engine = WiredTableRecognition() teds = TEDS() content = [] for one_data in test_data: img_path = one_data.",
+                content: "引言 link为了便于比较不同表格识别算法的效果差异，本篇文章基于评测工具TableRecognitionMetric和表格测试数据集liekkas/table_recognition上计算不同算法的TEDS指标。\n指标结果 link 方法 TEDS RapidTable 0.58786 lineless_table_rec 0.50054 wired_table_rec 0.63316 评测步骤 link1. 安装评测数据集和评测工具包 link pip install table_recognition_metric pip install modelscope==1.5.2 pip install rapidocr_onnxruntime==1.3.8 2. 安装表格识别推理库 link pip install rapid_table pip install lineless_table_rec pip install wired_table_rec 3. 推理代码 link info 完整评测代码，请移步Gist\nfrom modelscope.msdatasets import MsDataset from rapid_table import RapidTable from lineless_table_rec import LinelessTableRecognition from wired_table_rec import WiredTableRecognition from table_recognition_metric import TEDS test_data = MsDataset.load( \"table_recognition\", namespace=\"liekkas\", subset_name=\"default\", split=\"test\", ) # 这里依次更换不同算法实例即可 table_engine = RapidTable() # table_engine = LinelessTableRecognition() # table_engine = WiredTableRecognition() teds = TEDS() content = [] for one_data in test_data: img_path = one_data.get(\"image:FILE\") gt = one_data.get(\"label\") pred_str, _ = table_engine(img_path) scores = teds(gt, pred_str) content.append(scores) print(f\"{img_path}\\t{scores:.5f}\") avg = sum(content) / len(content) print(f'{avg:.5f}') 4. 写在最后 link以上评测仅是基于表格测试数据集liekkas/table_recognition测试而来，不能完全代表模型效果。\n因为每个模型训练数据不同，测试数据集如与训练数据相差较大，难免效果较差，请针对自身场景客观看待评测指标。\n"
+            }
+        );
+    index.add(
+            {
+                id:  9 ,
                 href: "\/TableStructureRec\/docs\/sponsor\/",
                 title: "给作者加油",
                 description: "写在前面 linkI like open source and AI technology because I think open source and AI will bring convenience and help to people in need, and will also make the world a better place. By donating to these projects, you can join me in making AI bring warmth and beauty to more people.\n我喜欢开源，喜欢AI技术，因为我认为开源和AI会为有需要的人带来方便和帮助，也会让这个世界变得更好。通过对这些项目的捐赠，您可以和我一道让AI为更多人带来温暖和美好。\n知识星球RapidAI私享群 link这里的提问会优先得到回答和支持，也会享受到RapidAI组织后续持续优质的服务，欢迎大家的加入。\n支付宝或微信打赏 (Alipay reward or WeChat reward) link通过支付宝或者微信给作者打赏，请写好备注。 Give the author a reward through Alipay or WeChat.",
@@ -848,7 +869,7 @@ <h3 id="参考资料">参考资料 <a href="#%e5%8f%82%e8%80%83%e8%b5%84%e6%96%9
         );
     index.add(
             {
-                id:  9 ,
+                id:  10 ,
                 href: "\/TableStructureRec\/docs\/faq\/",
                 title: "常见问题",
                 description: "",
@@ -857,7 +878,7 @@ <h3 id="参考资料">参考资料 <a href="#%e5%8f%82%e8%80%83%e8%b5%84%e6%96%9
         );
     index.add(
             {
-                id:  10 ,
+                id:  11 ,
                 href: "\/TableStructureRec\/docs\/changelog\/",
                 title: "更新日志",
                 description: "2023-11-22 update: link 发布wired_table_rec==0.0.1包 2023-11-20 update: link 发布lineless_table_rec==0.0.1包 2023-11-19 update: link 添加无线表格识别算法（LORE）的推理代码 ",
@@ -866,7 +887,7 @@ <h3 id="参考资料">参考资料 <a href="#%e5%8f%82%e8%80%83%e8%b5%84%e6%96%9
         );
     index.add(
             {
-                id:  11 ,
+                id:  12 ,
                 href: "\/TableStructureRec\/docs\/",
                 title: "Docs",
                 description: "",