🛠 Contribute: Found a typo? Or any other change in the description that you would like to see? Please consider sending us a pull request in the public repo of the challenge here.

🕵️ Introduction

We are taught to recognise words and letters as we grow up. And in some years, we start reading sentences, books, and more. How long does it take for an AI to start reading? Recognising written letters and words? Arguably, not as long as it takes us 😉

In this challenge, you are given a series of images with weird text on them, and you have to train an AI to identify what's written! This challenge aims to build a reliable model that extracts text from images. You will be provided with a dataset that contains 3 folders; training, validation & testing data. The task is to identify text on the image.

Understand with code! Here is getting started code for you.😄

💾 Dataset

The dataset contains 3 folders, training, validation & testing, the task is to identify text written on the image. The image is of size 256, 256 with text on different fonts with labels stored in the CSV file. The training dataset contains over 40000 images, validations has 10000 images and testing dataset contains 10000 images for predictions.

📁 Files

Following files are available in the resources section:

• train.csv - (40000 samples) This csv file contains the labels of the training images images.

• train.zip - The zip contains image for training set.

• val.csv - (4000 samples) This csv file contains the labels of the validation images.

• val.zip - The zip contains image for validation set.

• submission.csv - (10000 samples) This csv file is sample format of the submiting predictions of test images.

• test.zip - The zip contains image for testing set.

🚀 Submission

• Prepare a CSV containing image_id in sorted order and label as headers and the predicted text next to the image ids.
• Sample submission format available at sample_submission.csv in the resorces section.

Make your first submission here 🚀 !!

🖊 Evaluation Criteria

During evaluation Word Error Rate be used to test the efficiency of the model.

🔗 Links

• 💪 Challenge Page: https://www.aicrowd.com/challenges/txtocr
• 🗣️ Discussion Forum: https://www.aicrowd.com/challenges/txtocr/discussion
• 🏆 Leaderboard: https://www.aicrowd.com/challenges/txtocr/leaderboards

📱 Contact

• Shubhamai