Reference no: EM133369172
Scene-centric Image recognition task write code in python programming. Description is not required.
Required Databases
How to construct the training and testing dataset (Very important)
Question 1) For SUN397 datasets, please randomly split the datasets into a training dataset and a testing dataset, each with 50 images per class.
How to obtain top-1 accuracy/total training time for each given dataset.
Question 1) Any classification methods are allowed to use to obtain the average top-1 accuracy such as deep convolutional neural networks, RBF network, Extreme Learning Machine, Support Vector Machine, etc.
Question 2) feature extraction method(s) or feature fusion method(s) should be used to boost the performance such as autoencoder, deep believe networks, feature fusion, etc. Multiple models are encourage to use to generate heterogeneous feature for the final classifier.
Question 3) data argumentation algorithm(s) should be used to further boost the performance.
The training time should be included all the training time provided by the used classification methods, feature extraction methods, and data argumentation algorithms