Shanshan Wang
commited on
Commit
•
50fb3ca
1
Parent(s):
3a70a0e
added GOT-OCR2.0
Browse files- 0926-OCRBench-opensource.csv +3 -5
- app.py +1 -1
0926-OCRBench-opensource.csv
CHANGED
@@ -1,10 +1,6 @@
|
|
1 |
Rank,Model,Param (B),Language Model,Vision Model,OCRBench,Text Recognition,Scene Text-centric VQA,Document Oriented VQA,KIE,Handwritten Math Expression Recognition,
|
2 |
-
1,InternVL2-Llama3-76B,76,Llama-3-70B-Instruct,InternViT-6B,,,,,,,
|
3 |
-
2,InternVL2-40B,40,Nous-Hermes-2-Yi-34B,InternViT-6B,,,,,,,
|
4 |
-
3,InternVL2-26B,26,InternLM2-20B,InternViT-6B,,,,,,,
|
5 |
-
4,InternVL2-8B,8,InternLM2.5-7B,InternViT-300M,,,,,,,
|
6 |
,minicpm_v2.6,8,Qwen2-7B,SigLIP-400M,836,259,184,169,182,42,
|
7 |
-
,MiniMonkey,2,,,792,250,178,126,171,67,
|
8 |
7,GLM-4v-9B,9,glm-4-9b,EVA-02-5B,,,,,,,
|
9 |
8,CogVLM2-19B-Chat,19,Llama-3-8B-Instruct,EVA2-CLIP-E,,,,,,,
|
10 |
5,InternVL2-4B,4,Phi-3,InternViT-300M,785,236,170,154,158,67,
|
@@ -21,9 +17,11 @@ Rank,Model,Param (B),Language Model,Vision Model,OCRBench,Text Recognition,Scene
|
|
21 |
15,Mini-InternVL-Chat-4B-V1.5,4,Phi-3,InternViT-300M,640,193,160,146,135,6,
|
22 |
20,PaliGemma-3B-mix-448,3,Gemma-2B,SigLIP-400M,613,242,165,88,118,0,
|
23 |
21,MiniCPM-V-2,2.8,MiniCPM-2.4B,SigLIP-400M,596,243,168,100,85,0,
|
|
|
24 |
16,Phi-3-Vision,4.2,Phi-3,CLIP ViT-L/14,640,196,159,137,148,0,
|
25 |
,Internvl2-26B,26,,,823,251,184,153,168,67,
|
26 |
,Qwen2-VL-2B-Instruct,2.1,Qwen2-1.5B,ViT-600M,812,265,172,146,174,55,
|
|
|
27 |
,doctr-default,0.05,,,,177,,,,,
|
28 |
,doctr-v2m(best),0.05,,,,256,,,,,
|
29 |
,,,,,,,,,,,
|
|
|
1 |
Rank,Model,Param (B),Language Model,Vision Model,OCRBench,Text Recognition,Scene Text-centric VQA,Document Oriented VQA,KIE,Handwritten Math Expression Recognition,
|
|
|
|
|
|
|
|
|
2 |
,minicpm_v2.6,8,Qwen2-7B,SigLIP-400M,836,259,184,169,182,42,
|
3 |
+
,MiniMonkey,2.2,,,792,250,178,126,171,67,
|
4 |
7,GLM-4v-9B,9,glm-4-9b,EVA-02-5B,,,,,,,
|
5 |
8,CogVLM2-19B-Chat,19,Llama-3-8B-Instruct,EVA2-CLIP-E,,,,,,,
|
6 |
5,InternVL2-4B,4,Phi-3,InternViT-300M,785,236,170,154,158,67,
|
|
|
17 |
15,Mini-InternVL-Chat-4B-V1.5,4,Phi-3,InternViT-300M,640,193,160,146,135,6,
|
18 |
20,PaliGemma-3B-mix-448,3,Gemma-2B,SigLIP-400M,613,242,165,88,118,0,
|
19 |
21,MiniCPM-V-2,2.8,MiniCPM-2.4B,SigLIP-400M,596,243,168,100,85,0,
|
20 |
+
21,MiniCPM-V,3,MiniCPM-2.4B,SigLIP-400M,596,243,168,100,85,0,
|
21 |
16,Phi-3-Vision,4.2,Phi-3,CLIP ViT-L/14,640,196,159,137,148,0,
|
22 |
,Internvl2-26B,26,,,823,251,184,153,168,67,
|
23 |
,Qwen2-VL-2B-Instruct,2.1,Qwen2-1.5B,ViT-600M,812,265,172,146,174,55,
|
24 |
+
,GOT-OCR2.0,0.6,,,622,245,99,83,164,31,
|
25 |
,doctr-default,0.05,,,,177,,,,,
|
26 |
,doctr-v2m(best),0.05,,,,256,,,,,
|
27 |
,,,,,,,,,,,
|
app.py
CHANGED
@@ -19,7 +19,7 @@ dtype_dict = {
|
|
19 |
|
20 |
|
21 |
# preprocess the dataframe
|
22 |
-
data_valid = data[:
|
23 |
data_valid = data_valid.astype(dtype_dict)
|
24 |
data_valid.drop(columns=['Unnamed: 11'], inplace=True)
|
25 |
|
|
|
19 |
|
20 |
|
21 |
# preprocess the dataframe
|
22 |
+
data_valid = data[:25].copy()
|
23 |
data_valid = data_valid.astype(dtype_dict)
|
24 |
data_valid.drop(columns=['Unnamed: 11'], inplace=True)
|
25 |
|