alkzar90/CC6204-Hackaton-Cub-Dataset

Dataset Overview

Basic Information

• Dataset Name: CC6204‑Hackaton‑CUB200
• License: Apache‑2.0
• Language: English
• Size Category: 10K<n<15K
• Source Dataset: Extension | Other
• Task Categories:
  • Image Classification
  • Text Classification
• Task ID: Multiclass Image Classification
• Paper/Code ID: cub‑200‑2011

Dataset Description

• Homepage: CUB 200 2011
• Repository: Caltech Vision Lab
• Paper: The Caltech‑UCSD Birds‑200‑2011 Dataset
• Contact: Catherine Wah

Data Instances

• Image: RGB image representing a bird
• Description: Ten textual captions per image, separated by newline characters
• Label: Integer representing the bird species ID
• Filename: Image file name

Data Splits

• Training Set: 5,994 observations
• Test Set: 5,794 observations

Problem Statement

The goal is to train models to achieve optimal classification of CUB instances. Experiments may explore image‑only, text‑only, or combined multimodal approaches.

Experimental Strategy

Given limited compute resources, a few‑shot strategy is recommended, e.g., reducing per‑class samples or limiting the number of classes.

Evaluation Metric

Accuracy on the test set.

Citation Information

@techreport{WahCUB_200_2011,
	Title = {The Caltech‑UCSD Birds‑200‑2011 Dataset},
	Author = {Wah, C. and Branson, S. and Welinder, P. and Perona, P. and Belongie, S.},
	Year = {2011},
	Institution = {California Institute of Technology},
	Number = {CNS‑TR‑2011‑001}
}