##Object-Recognition-on-aYahoo-dataset Object recognition algorithm to classify 12 categories of aYahoo Image dataset trained in Caffe framework.
###Categories Total Images- 2152
Class | Train Images | Test Images |
---|---|---|
Bag | 222 | 71 |
Building | 174 | 52 |
Carriage | 108 | 46 |
Centaur | 38 | 13 |
Donkey | 103 | 31 |
Goat | 118 | 39 |
Jetski | 134 | 41 |
Monkey | 129 | 36 |
Mug | 177 | 50 |
Statue | 163 | 36 |
Wolf | 142 | 41 |
Zebra | 144 | 43 |
All Classes | 1652 | 500 |
###Results ####Using alexnet model Accuracy in percentage obtained by taking various features of images and various training methods while using alexnet model.
alexnet | Linear Kernel | RBF Kernel | Polynomial Kernel | LinearSVC |
---|---|---|---|---|
First 500 features of fc7 layer | 91.6 | 35.6 | 84.4 | 90.2 |
First 2048 features of fc7 layer | 93.8 | 35.6 | 87.8 | 93.4 |
All features of fc7 layer | 94.4 | 35.6 | 88.4 | 94.2 |
First 500 features of fc8 layer | 92.8 | 35.6 | 91.6 | 92.6 |
All features of fc8 layer | 92.2 | 35.6 | 92.2 | 92.6 |
All features of fc7+fc8 layer | 93.4 | 35.6 | 92.8 | 93.8 |
####Using caffenet model Accuracy in percentage obtained by taking various features of images and various training methods while using caffenet model.
caffenet | Linear Kernel | RBF Kernel | Polynomial Kernel | LinearSVC |
---|---|---|---|---|
First 500 features of fc7 layer | 90.2 | 35.6 | 84.4 | 90.2 |
First 2048 features of fc7 layer | 92.8 | 35.6 | 87.8 | 93.4 |
All features of fc7 layer | 93.4 | 35.6 | 88.4 | 94.2 |
First 500 features of fc8 layer | 90.0 | 35.6 | 91.6 | 92.6 |
All features of fc8 layer | 90.6 | 35.6 | 92.2 | 92.6 |
All features of fc7+fc8 layer | 91.4 | 35.6 | 92.8 | 93.8 |