Deep Learning: Cat Meme Detector Architecture

Overview

A deep cat meme detector is a neural network that identifies and categorizes cat memes in a given image dataset. Our architecture consists of the following stages:

Preprocessing: The input image is resized to 256x256 and converted to grayscale.
Feature Extraction: SqueezeNet is used for feature extraction, which outputs 128x128x128 feature maps.
Convolutional Layer: 5 Conv2D layers are applied with kernel size 3x3 and ReLU activation.
Pooling Layer: Max pooling with a 2x2 kernel and stride 2.
Flatten: Flattening the output of the previous layer.
Fully Connected Layer: 2 FC layers with 512 units and ReLU activation.
Output Layer: Softmax output.

Components

Our deep cat meme detector consists of:

Input Image: The input image is a 256x256 RGB image.
Conv2D Layers: 5 Conv2D layers with 32, 64, 128, 256, and 512 filters.
Max Pooling Layers: 5 Max Pooling layers with 2x2 kernel and stride 2.
Flatten: Flattening the output of the previous layer.
Fully Connected Layers: 2 FC layers with 512 units and ReLU activation.

Parameter	Value
Learning Rate	0.01
Batch Size	32
Epochs	10

Deep Cat Meme Detector Architecture

Overview

Components

Hyperparameters