Changelog
September 25, 2023 ยท View on GitHub
v0.6.1 (25/09/2023)
Highlights
- Release the implements of CTUNet model and ComFinTab dataset. (ACMMM 2022)
v0.6.0 (13/07/2022)
Highlights
- Release the implements of DLD model (ECCV 2022)
v0.5.2 (13/07/2022)
Highlights
- Support for setting logging interval.
- Support for filtering invalid box in davar_loading.
- Support for multi-dataset testing in recognizor.
- Some bug fixed
v0.5.1 (12/05/2022)
Highlights
- Support for the Reading Order Detection task.
- Release re-implement of GCN-PN model (ECCV 2020).
v0.5.0 (08/05/2022)
Highlights
- Support for the basic NLP tasks, currently including NER. In the future we will release more works related to the text understanding.
- Release BERT-based NER models, including BERT+CRF, BERT+Softmax, BERT+Span.
- Release BiLSTM-CRF-based NER models.
- Provide the demo for table recognition with OCR (MaskRCNN + ATT + LGPMA).
v0.4.1 (04/05/2022)
Highlights
- Release the implementations of Chargrid.
- Some bug fixed.
v0.4.0 (31/12/2021)
Highlights
- Release the implementations of VSR.
v0.3.1 (20/12/2021)
Highlights
- Rebuild the directory architecture to fit with mmdetection.
- Format some of the paths
- Change all models/datalists' links into Hikvision One, which adapts for people anywhere.
v0.3.0 (22/09/2021)
Highlights
- Add the support for table recognition.
- Release the implementations of LGPMA.
v0.2.1 (16/09/2021)
Highlights
- Add the end-to-end model of Text-Perceptron-Spot.
- Update end-to-end model of Mask-RCNN-Spot.
v0.2.0 (24/08/2021)
Highlights
- Add support for video text task, and release the implementations of YORO.
- Some bug fixes.
v0.1.0 (23/07/2021)
Highlights
- Release DavarOCR open source OCR framework, which is compatible with mmdetection-2.11.0 and mmcv-1.3.4.
- Release the OCR algorithms published by Davar-Lab and some re-implementations of others algorithms.
v0.0.1 (15/12/2020)
- hikopensource setup, which is compatible with mmdetection-1.2.0 and mmcv-0.4.2.