Community Round: Completed #educational Weight: 25.0
3203
23
0
0

Note: Update in the Evaluator

π  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.

## π΅οΈ Introduction

This problem we are levelling up from using images to using videos are input! Youβve successfully identified the board position from an image. But can you predict moves of chess pieces from a video clip?

The input will be a short video of a game of chess, with pieces being moved around. Can you create a model that will identify from which location pieces moves and landed where? This is going to be an exciting challenge!

Understand with code! Here is getting started code for you.π

## πΎ Dataset

The given dataset contains videos of chess board with pieces moving around the board. Each frame in video is of size 256 * 256 width & hight. A CSV is also provided containing the VideoID and the previous & new locations of multiples pieces from the video. The format of moves are determined by the THE Cartesian Coordinate System and the space between in each label ( for ex. b2b3 d8f7 ) is describing move from a different piece.

Sample Column & Video

|VideoID|label |
|-------|---------------------------------------------|
|0      |b2b3 d8f7 f4f5 c2c1r h6f4 f7d8 f2f3 c1b1 b3b7|

The dataset is divided into train and validation set, each containing a zip file and csv corresponding to it. For evaluation you are provided with the test.zip which contain the videos for which you need to find the previous & new locations of pieces.

## π Files

Following files are available in the resources section:

• train.csv - (5000 samples) This csv file contains VideoID column which corresponds to train.zip and labels
• train.zip - (5000 samples) This zip file contains video corresponding to the first column of train.csv.
• val.csv - (1000 samples) This csv file contains VideoID column which corresponds to val.zip and labels
• val.zip - (1000 samples) This zip file contains video corresponding to the first column of val.csv.

• test.zip - (2000 samples) This zip file contains testing videos that will be used for actual evaluation for the leaderboard score.

## π Submission

• Prepare a CSV containing headers as VideoID and label containing predicted moves.
• Sample submission format available at sample_submission.csv in the resorces section.

Make your first submission here π !!

## π Evaluation Criteria

During evaluation Word Error Rate will be used to test the efficiency of the model and in python using jiwer. For ex.

from jiwer import wer

ground_truth = 'hello world' hypothesis = 'hello duck'

error = wer(ground_truth, hypothesis)