Watch, Attend and Parse for Handwritten Mathematical Expression Recognition

Implementation of various Vision Transformers I found interesting

Adversarial Generation of Handwritten Text Images

E2E-MLT - an Unconstrained End-to-End Method for Multi-Language Scene Text

Unsupervised Intra-domain Adaptation for Semantic Segmentation through Self-Supervision (CVPR 2020 Oral)

The scripts for training Detectron2-based Layout Models on popular layout analysis datasets

ScrabbleGAN: Semi-Supervised Varying Length Handwritten Text Generation (CVPR20)

Official code for the paper "DeepSVG: A Hierarchical Generative Network for Vector Graphics Animation". Includes a PyTorch library for deep learning with SVG data.

PDFTableExtract

Bezier AE approach to sketch generation

Code for the paper "VirTex: Learning Visual Representations from Textual Annotations"

[CVIU 2019] ASSD learns to highlight useful regions on the feature maps while suppressing the irrelevant information, thereby providing reliable guidance for object detection.

Ground truth line annotations for the Berliner Börsen-Zeitung

SUPERT: Unsupervised multi-document summarization evaluation & generation

[WACV2021] Oriented Object Detection in Aerial Images with Box Boundary-Aware Vectors