Colored MNIST Dataset A comprehensive dataset of MNIST digits with RGB colored backgrounds, designed for multi objective classification tasks including digit recognition and background color classification. Overview This dataset extends the classic MNIST digit dataset by adding colored backgrounds (blue, green, red) to create a multi label classification challenge. Each image contains: Digit label : 0 9 (digit recognition) Color label : blue, green, red (background color classification) Image format : 28×28 RGB PNG files Dataset Statistics Training Set (8,835 samples) Digit Count Digit Count 0 868 5 764 1 1,000 6 893 2 858 7 948 3 909 8 835 4 883 9 877 Test Set (10,000 samples) Digit Count Digit Count 0 980 5 892 1 1,135 6 958 2 1,032 7 1,028 3 1,010 8 974 4 982 9 1,009 Color Distribution Blue : ~2,946 training samples Green : ~2,968 training samples Red : ~2,921 training samples Directory Structure Usage Loading with HuggingFace Datasets Applications This dataset is ideal for: 1. Multi task Learning : Train models to simultaneously classify digits and colors 2. Domain Adaptation : Study how background colors affect digit recognition 3. Bias Analysis : Investigate color digit corre…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy