- Task 1 - Detection
- Task 2 - Detection-Linking
- Task 3 - Detection-Recognition
- Task 4 - Detection-Recognition-Linking
method: Baseline TESTR Checkpoint2024-03-26
Authors: Organizers
Affiliation: ICDAR'24 RRC-MapText
Description: TESTR checkpoint is used without any additional modifications or finetuning. The model checkpoint version with polygon prediction head and fine-tuned on TotalText was used.
method: MapText Using EasyOCR2024-05-04
Authors: Pengyu Chen; Xuezi Bi
Affiliation: University of South Carolina; Sun Yat-sen University
Description: EasyOCR is a python module for extracting text from image. It is a general OCR that can read both natural scene text and dense text in document.
We chose EasyOCR as our model and fine-tuned it for this detection task. This is an end-to-end model.
1. due to GPU limitation (we are using standard colab), we don't have enough epochs for training.
2. due to time constraints (we just learned about this contest in April), we didn't have enough fine-tuning experiments.
I hope you can inform me about better model or better methods after the contest, I am very interested in these tasks.
| Date | Method | Quality | F-score | Tightness | Precision | Recall | |||
|---|---|---|---|---|---|---|---|---|---|
| 2024-03-26 | Baseline TESTR Checkpoint | 55.13% | 69.29% | 79.57% | 71.85% | 66.90% | |||
| 2024-05-04 | MapText Using EasyOCR | 42.67% | 58.33% | 73.16% | 69.29% | 50.36% |