How could we combine ANN+CNN and combining CNN+SVM?

Question

I have trained SVM ,CNN and ANN on UCF-101 dataset,SVM and ANN uses a CSV files of Hue and LBP features respectively, while CNN uses LBP images for classification training.Now I want to combine {SVM and CNN} and {ANN and CNN}.Is it possible to do so? if yes, how can it be done.

I have extracted 1st Keyframe for every video in the Dataset and then calculated its LBP histogram.Used it as feature for the image ,wrote it in csv along with its label(I have selected data of only 5 classes out of 101 available) and then trained SVM and ANN on it.ANN is simple network with Softmax Activation function at the output and 2 hidden layers with 8 nodes in each hidden layer. No dropout is used. For CNN, I extracted LBP features of the image and converted it to image . This image data is what i used to train CNN. CNN network has input layer,convolution layer,activation layer,full connection layer,dropout layer(dropout=0.5) and output layer .Activation function is softmax

SVM gives 10% accuracy in testing while ANN gives 35% for 1000 epoch. CNN gives accuracy of 29% for test data with 10 epoch

Answer 1

First of all, you need to increase hidden layer or hidden nodes because your accuracy for CNN, ANN, SVM is not good. Also you don`t need to add Dropout layer always, especially in this situation whereby your accuracy is not good. Originally, Dropout layer is for avoiding overfitting.

Secondly, I don`t know how your SVM,CNN,ANN code is suppose to look like. However, according to you saying that, there is probability that each SVM, CNN, ANN code is a little bit wrong.

So check your code again.