Image to Latex

Introduction

This respository implement the Seq2Seq Image to Latex architecture from paper “Image to Latex.” of Genthial, Guillaume. (2017).

Architecture

This structure is based on Seq2Seq architecture, it use one Convolutional Encoder and one RNN Decoder.

• Convolution (only)
• Convolution with Row Encoder (BiLSTM)
• Convolution with Batch Norm
• ResNet 18 with Row Encoder (BiLSTM)
• ResNet 18 (only)

Dataset

im2latex100k

• https://www.kaggle.com/datasets/shahrukhkhan/im2latex100k
• https://www.kaggle.com/datasets/tuannguyenvananh/im2latex-sorted-by-size

im2latex170k

• https://www.kaggle.com/datasets/rvente/im2latex170k
• https://www.kaggle.com/datasets/tuannguyenvananh/im2latex-170k-meta-data

How to use?

Login wandb

• wandb login <key>

python main.py --batch-size 2 --data-path C:\Users\nvatu\OneDrive\Desktop\dataset5\dataset5 --img-path C:\Users\nvatu\OneDrive\Desktop\dataset5\dataset5\formula_images --dataset 170k --val --decode-type beamsearch

Example

• ResNet Row Encoder