echo840 commited on
Commit
1e094a2
β€’
1 Parent(s): 6c79346

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +104 -6
app.py CHANGED
@@ -54,7 +54,7 @@ def load_leaderboard_table_csv(filename, add_hyperlink=True):
54
  for j in range(len(heads)):
55
  item = {}
56
  for h, v in zip(heads, row):
57
- if h != "Model" and h != "Link":
58
  item[h] = int(v)
59
  else:
60
  item[h] = v
@@ -76,6 +76,12 @@ def get_arena_table(model_table_df):
76
  # model display name
77
  row.append(model_name)
78
 
 
 
 
 
 
 
79
  row.append(
80
  model_table_df["Text Recognition"].values[model_key]
81
  )
@@ -102,10 +108,59 @@ def get_arena_table(model_table_df):
102
  values.append(row)
103
  return values
104
 
105
- def build_leaderboard_tab(leaderboard_table_file, show_plot=False):
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
106
  if leaderboard_table_file:
107
  data = load_leaderboard_table_csv(leaderboard_table_file)
 
108
  model_table_df = pd.DataFrame(data)
 
109
  md_head = f"""
110
  # πŸ† OCRBench Leaderboard
111
  | [GitHub](https://github.com/Yuliang-Liu/MultimodalOCR) | [Paper](https://arxiv.org/abs/2305.07895) |
@@ -121,6 +176,8 @@ def build_leaderboard_tab(leaderboard_table_file, show_plot=False):
121
  headers=[
122
  "Rank",
123
  "Name",
 
 
124
  "Text Recognition",
125
  "Scene Text-Centric VQA",
126
  "Doc-Oriented VQA",
@@ -131,6 +188,45 @@ def build_leaderboard_tab(leaderboard_table_file, show_plot=False):
131
  datatype=[
132
  "str",
133
  "markdown",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
134
  "number",
135
  "number",
136
  "number",
@@ -141,9 +237,10 @@ def build_leaderboard_tab(leaderboard_table_file, show_plot=False):
141
  value=arena_table_vals,
142
  elem_id="arena_leaderboard_dataframe",
143
  height=700,
144
- column_widths=[60, 120, 150, 200, 180, 80, 80, 160],
145
  wrap=True,
146
  )
 
147
  else:
148
  pass
149
  md_tail = f"""
@@ -151,7 +248,7 @@ def build_leaderboard_tab(leaderboard_table_file, show_plot=False):
151
  If you would like to include your model in the OCRBench leaderboard, please follow the evaluation instructions provided on [GitHub](https://github.com/Yuliang-Liu/MultimodalOCR) and feel free to contact us via email at zhangli123@hust.edu.cn. We will update the leaderboard in time."""
152
  gr.Markdown(md_tail, elem_id="leaderboard_markdown")
153
 
154
- def build_demo(leaderboard_table_file):
155
  text_size = gr.themes.sizes.text_lg
156
 
157
  with gr.Blocks(
@@ -160,7 +257,7 @@ def build_demo(leaderboard_table_file):
160
  css=block_css,
161
  ) as demo:
162
  leader_components = build_leaderboard_tab(
163
- leaderboard_table_file, show_plot=True
164
  )
165
  return demo
166
 
@@ -168,7 +265,8 @@ if __name__ == "__main__":
168
  parser = argparse.ArgumentParser()
169
  parser.add_argument("--share", action="store_true")
170
  parser.add_argument("--OCRBench_file", type=str, default="./OCRBench.csv")
 
171
  args = parser.parse_args()
172
 
173
- demo = build_demo(args.OCRBench_file)
174
  demo.launch()
 
54
  for j in range(len(heads)):
55
  item = {}
56
  for h, v in zip(heads, row):
57
+ if h != "Model" and h != "Link" and h != "Language Model" and h != "Open Source":
58
  item[h] = int(v)
59
  else:
60
  item[h] = v
 
76
  # model display name
77
  row.append(model_name)
78
 
79
+ row.append(
80
+ model_table_df["Language Model"].values[model_key]
81
+ )
82
+ row.append(
83
+ model_table_df["Open Source"].values[model_key]
84
+ )
85
  row.append(
86
  model_table_df["Text Recognition"].values[model_key]
87
  )
 
108
  values.append(row)
109
  return values
110
 
111
+ def get_recog_table(model_table_df):
112
+ # sort by rating
113
+ values = []
114
+ for i in range(len(model_table_df)):
115
+ row = []
116
+ model_key = model_table_df.index[i]
117
+ model_name = model_table_df["Model"].values[model_key]
118
+ # rank
119
+ row.append(i + 1)
120
+ # model display name
121
+ row.append(model_name)
122
+
123
+ row.append(
124
+ model_table_df["Language Model"].values[model_key]
125
+ )
126
+ row.append(
127
+ model_table_df["Open Source"].values[model_key]
128
+ )
129
+ row.append(
130
+ model_table_df["Regular Text"].values[model_key]
131
+ )
132
+
133
+ row.append(
134
+ model_table_df["Irregular Text"].values[model_key]
135
+ )
136
+
137
+ row.append(
138
+ model_table_df["Artistic Text"].values[model_key]
139
+ )
140
+
141
+ row.append(
142
+ model_table_df["Handwriting"].values[model_key]
143
+ )
144
+
145
+ row.append(
146
+ model_table_df["Digit string"].values[model_key]
147
+ )
148
+
149
+ row.append(
150
+ model_table_df["Non-semantic Text"].values[model_key]
151
+ )
152
+ row.append(
153
+ model_table_df["ALL"].values[model_key]
154
+ )
155
+ values.append(row)
156
+ return values
157
+
158
+ def build_leaderboard_tab(leaderboard_table_file, text_recog_file, show_plot=False):
159
  if leaderboard_table_file:
160
  data = load_leaderboard_table_csv(leaderboard_table_file)
161
+ data_recog = load_leaderboard_table_csv(text_recog_file)
162
  model_table_df = pd.DataFrame(data)
163
+ recog_table_df = pd.DataFrame(data_recog)
164
  md_head = f"""
165
  # πŸ† OCRBench Leaderboard
166
  | [GitHub](https://github.com/Yuliang-Liu/MultimodalOCR) | [Paper](https://arxiv.org/abs/2305.07895) |
 
176
  headers=[
177
  "Rank",
178
  "Name",
179
+ "Language Model",
180
+ "Open Source",
181
  "Text Recognition",
182
  "Scene Text-Centric VQA",
183
  "Doc-Oriented VQA",
 
188
  datatype=[
189
  "str",
190
  "markdown",
191
+ "str",
192
+ "str",
193
+ "number",
194
+ "number",
195
+ "number",
196
+ "number",
197
+ "number",
198
+ "number",
199
+ ],
200
+ value=arena_table_vals,
201
+ elem_id="arena_leaderboard_dataframe",
202
+ height=700,
203
+ column_widths=[60, 120,150,100, 150, 200, 180, 80, 80, 160],
204
+ wrap=True,
205
+ )
206
+ with gr.Tab("Text Recognition", id=1):
207
+ arena_table_vals = get_recog_table(recog_table_df)
208
+ md = "OCRBench is a comprehensive evaluation benchmark designed to assess the OCR capabilities of Large Multimodal Models. It comprises five components: Text Recognition, SceneText-Centric VQA, Document-Oriented VQA, Key Information Extraction, and Handwritten Mathematical Expression Recognition. The benchmark includes 1000 question-answer pairs, and all the answers undergo manual verification and correction to ensure a more precise evaluation."
209
+ gr.Markdown(md, elem_id="leaderboard_markdown")
210
+ gr.Dataframe(
211
+ headers=[
212
+ "Rank",
213
+ "Name",
214
+ "Language Model",
215
+ "Open Source",
216
+ "Regular Text",
217
+ "Irregular Text",
218
+ "Artistic Text",
219
+ "Handwriting",
220
+ "Digit string",
221
+ "Non-semantic Text",
222
+ "ALL",
223
+ ],
224
+ datatype=[
225
+ "str",
226
+ "markdown",
227
+ "str",
228
+ "str",
229
+ "number",
230
  "number",
231
  "number",
232
  "number",
 
237
  value=arena_table_vals,
238
  elem_id="arena_leaderboard_dataframe",
239
  height=700,
240
+ column_widths=[60, 120,150,100, 100, 100, 100, 100, 100,100, 80],
241
  wrap=True,
242
  )
243
+
244
  else:
245
  pass
246
  md_tail = f"""
 
248
  If you would like to include your model in the OCRBench leaderboard, please follow the evaluation instructions provided on [GitHub](https://github.com/Yuliang-Liu/MultimodalOCR) and feel free to contact us via email at zhangli123@hust.edu.cn. We will update the leaderboard in time."""
249
  gr.Markdown(md_tail, elem_id="leaderboard_markdown")
250
 
251
+ def build_demo(leaderboard_table_file, recog_table_file):
252
  text_size = gr.themes.sizes.text_lg
253
 
254
  with gr.Blocks(
 
257
  css=block_css,
258
  ) as demo:
259
  leader_components = build_leaderboard_tab(
260
+ leaderboard_table_file, recog_table_file,show_plot=True
261
  )
262
  return demo
263
 
 
265
  parser = argparse.ArgumentParser()
266
  parser.add_argument("--share", action="store_true")
267
  parser.add_argument("--OCRBench_file", type=str, default="./OCRBench.csv")
268
+ parser.add_argument("--TextRecognition_file", type=str, default="./TextRecognition.csv")
269
  args = parser.parse_args()
270
 
271
+ demo = build_demo(args.OCRBench_file, args.TextRecognition_file)
272
  demo.launch()