# Neural Network Basics 2

1.  Input Layer -
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753467269671/8e3a7611-ffe8-4dda-a1d2-1aa30964ab78.png align="left")
    
    Drawn image ko pixels me convert karta hai
    
    *   Raw input in a neural network
        
2.  Convolutional Layer -
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753466174008/1f181707-566f-46ae-8d43-f6cf3197b2dd.png align="left")
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753466228596/d73b8711-1b96-4d72-850b-541319fac149.png align="left")
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753466290096/32989b60-76f7-4542-a058-7b1817d1b632.png align="left")
    
    Multiple layers create hoti hai by a single layer
    
    *   har layer alag alag views show karti hai like different patterns and all
        
    *   Stride - 1
        
    *   For finding patterns, various filters are also applied in this layer, such as edge detection and curve detection.
        
3.  Downsampling Layer -
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753467608943/1130dceb-51c8-410d-bd24-6f3d8cc4fdbc.png align="center")
    
    Pixel reduce karta hai
    
    *   Stride - 2
        
    *   This is also known as a Pooling Layer
        
    *   Is layer see speed increase hoti hai, and overfitting reduce hoti hai
        
4.  Flatten Layer
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753468468322/b07f9f8d-c6b9-4873-8483-211f6adafcdc.png align="center")
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753468668788/dd3e7866-c256-4a55-ad11-78be54cd9aae.png align="center")
    
    is layer me, image gets converted into a complete vector
    
    *   3D ya 2D layer se 1D Line me convert ho jaata hai
        
5.  Dense Layer
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753469952070/412dc3d2-362e-4a8a-8436-a95562fd48b5.png align="center")
    
    Flatten Layer jaisa hi rhta hai but ye convert nahi karta… ye 1d vector ko kam karta hai(matlab dense karta hai)
    
    *   jaise 128 se 64 me convert karna
        
6.  Output Layer
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753470094695/00aee172-9264-4aa3-b81c-39a44c328732.png align="center")
    
    Ye bhi ek dense layer hoti hai but iska ek specific goal hota hai
    
    *   ye batata hai final output(final value/final prediction)