| MLP | CNN |
|---|---|
| Only use fully-connected layers | Makes use of fully as well as sparsely connected layers |
| Only accepts vector as input | Also accepts matrices as input |
In a MLP, every hidden node needs to be connected to every pixel in input image. This accounts for a lot of redundancy.
Instead of every node keeping information of every pixel of input image, we divide the image into 4 regions - red, green, blue and yellow. Then each hidden node can be connected to only the pixels in one of these 4 regions as depicted below.
To install opencv follow either of the below 2 methods -
- Install OpenCV on Linux
- Install OpenCV on Windows
- If you have Anaconda Python installed, just run the below command -
conda install opencv
Note - You can validate installation of opencv by running, import opencv