Painter classification using genetic algorithms

Abstract

Abstract-This paper describes the problem of painter classification. We propose solving the problem by using genetic algorithms, which yields very promising results. The proposed methodology combines dimensionality reduction (via image preprocessing) and evolutionary computation techniques, by representing preprocessed data as a chromosome for a genetic algorithm (GA). The preprocessing of our scheme incorporates a diverse set of complex features (e.g., fractal dimension, Fourier spectra coefficients, and texture). The training phase of the GA employs a weighted nearest neighbor (NN) algorithm. We provide initial promising results for the 2-and 3-class cases, which offer significant improvement in comparison to a standard nearest neighbor classifier.

A Survey of Genetic Algorithms for Multi-Label Classification

“… It is also important to observe that over the last decades, GAs have been successfully used to solve a large number of SLC problems in very distinct contexts and application domains. For instance, GAs have been widely employed to perform feature selection [2], [43]- [46], to determine the best set of weights for training neural networks [47] and to discover classification rules [38], [43], [48].…”

DeepPainter: Painter Classification Using Deep Convolutional Autoencoders

“…The supervised classification benchmark is identical to that used by Levy et al [9,10] in their experiments. It consists of (3×40 =) 120 digital reproductions of paintings by Rembrandt, Renoir, and van Gough, downloaded from the Webmuseum.…”

“…the input layer consists of the raw image (resampled to 256 × 256 pixels) in three channels (R, G, and B) 2. convolutional layer with 100 5 × 5 filters per input channel 3. max-pooling layer of size 2 × To make our results directly comparable to those of Levy et al [9,10], we conducted 10-fold cross validation, where in each of 10 runs 90% of the data is used for training, and 10% for validation.…”

“…Levy et al [9] used GA-based WNN with a set of 78 prevalent image features for classifying paintings by Rembrandt, Renoir, and van Gogh, obtaining 80% classification accuracy. In their later work [10], they augmented their approach by also adding 20 features using restricted Boltzmann machines (RBM) [2], improving the classification accuracy to 90.44%.…”

“…Levy et al [9,10] applied feature extraction to paintings using generic image processing (IP) functions (e.g., fractal dimension, Fourier spectra coefficients, texture coefficients, etc. ), and restricted Boltzmann machines (RBM), followed by genetic algorithms (GA)-based learning of the weights of a weighted nearest neighbor (WNN) classifier [13].…”