FinTab model is less accurate when it comes to financial reports?

microsoft / table-transformer

Table Transformer (TATR) is a deep learning model for extracting tables from unstructured documents (PDFs and images). This is also the official repository for the PubTables-1M dataset and GriTS evaluation metric.

MIT License

2.01k stars 231 forks source link

FinTab model is less accurate when it comes to financial reports? #143

Open linkstatic12 opened 9 months ago

linkstatic12 commented 9 months ago

After comparing TATR-v1.1-Fin-msft.pth, TATR-v1.1-All-msft.pth and nielsr/detr-table-structure-recognition I found out that nielsr/detr-table-structure-recognition is more accurate when it comes to financial reports structure recognition. Am I doing something wrong?

linkstatic12 commented 9 months ago

pipe= TableExtractionPipeline(det_config_path='detection_config.json', det_model_path='../pubtables1m_detection_detr_r18.pth', det_device='cuda', str_config_path='structure_config.json', str_model_path='../TATR-v1.1-Fin-msft.pth', str_device='cuda') example of usage of fin tab weights only