Authors: Needa Fatima Attar, Sana Razeen, Triveni Kalwad

Affiliation: KLE Technological University

Email: 01fe20bcs184@kletech.ac.in

Description: The subtasks of detection and recognition on roadtext1k[1] dataset is performed using EasyOCR, which uses CRAFT[2] algorithm for detection and CRNN[3] as its recognition model. After the framewise detection and recognition of each video, the subtask of tracking is performed using the unique transcriptions of each video. Each unique text transcription of a video gets a specific ID assigned. The repeated instances of a unique text transcription get same ID throughout the video.


[1] Reddy, Sangeeth, et al. "Roadtext-1k: Text detection & recognition dataset for driving videos." 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020.

[2] Baek, Youngmin, et al. "Character region awareness for text detection." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019.

[3]Shi, Baoguang, Xiang Bai, and Cong Yao. "An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition." IEEE transactions on pattern analysis and machine intelligence 39.11 (2016): 2298-2304.