Chess Piece Identification

Chess Piece Identification

Chess is a famous strategic game with a centuries-old history, recognized for its complexity and high demands for logical thinking and planning. The connection between chess and artificial intelligence is close: many algorithms and machine learning techniques have been developed precisely to create agents capable of playing chess at a high level.

A historic moment in this context was in 1997, when the Deep Blue computer, developed by IBM, defeated world champion Garry Kasparov, demonstrating the potential of computers in complex strategic games.

---

🗂 Data provided

You have at your disposal a dataset composed of the following files:

📁 train.csv

Contains:

• id – unique identifier (P00001, P00002 …)
• image_path – path to image
• label – piece type (bishop, knight, pawn, queen, rook)

📁 test.csv

Contains:

• id – unique identifier
• image_path – path to image

All images are stored in the images/ directory.

---

🎯 Task

Build an image classification model that can differentiate chess pieces (bishop, knight, pawn, queen, rook). You can use any modern deep learning architecture: ResNet, EfficientNet, Vision Transformer, ConvNeXt etc., possibly with fine-tuning on pre-trained models.

At the end, you will generate a submission.csv file with the format:

id,label
P00081,queen
P00082,knight

---

📊 Evaluation

The metric used for model evaluation is accuracy:

accuracy = (number_of_correct_predictions / total_number_of_predictions)

🏅 Scoring system:

• accuracy ≥ 90% → 100 points
• accuracy < 20% → 0 points
• otherwise → proportional score between 0 and 100