Hi! myself Siddharth I have made a simple classification project using the TensorFlow library
Images has be downloaded from Kaggle.com (link to image dataset) ==> Cats-vs-Dog Dataset
We then upzip the folder and then arrange the files in order accordingly creating the following folders having the files
-
tmps
-
imagetest
-
tmps/cats-v-dogs\training
-
tmps/cats-v-dogs\validation
-
tmps/cats-v-dogs\training\cats
-
tmps/cats-v-dogs\training\dogs
-
tmps/cats-v-dogs\validation\cats
-
tmps/cats-v-dogs\validation\dogs
The imagetest folder consist of images which u can download from WEB
- The model uses Convolution to find out important features of Data To know more medium link About Convolution and Its derivation
- The first layer, Conv2D, is a convolutional layer with 3x3 filters and ReLU activation function.
- It takes in 150 x 150 pixel input images and outputs the result as 3 x 3 feature maps.
- MaxPooling2D, is a max pooling layer with 2x2 filters and ReLU activation function.
- It takes in 2 x 2 feature maps from the previous layer and outputs one final feature map with size of (150, 150).
- Repeating the same for 3 times ( my cpu is suffering from strokes 😂😂)
- Then we Flattern it
- Finally now model is ready to get trained
- After the model gets trained we can now test it using custom inputs which u can input
The model can be fine tuned much more here are few things u can do
- Changing the Image Augmentation like
- Changing the Dimensions of the image
- Cropping the Image
- Shearing the Image
- Adding Noise
- Adding more Convolutional Layers
- Adding Droupouts (set it to 0.3 max)
If anyone out there in the community knows how to use GPU's in laptop/computers please help me to use the GPU and also help me how to use GPU parallelization to run the models even more faster
GPU SPEC
- Currently using a Laptop GPU (GTX 1650 TI)