What computer vision encompasses - Model Pipeline Trace

Model Pipeline - What computer vision encompasses

Computer vision helps computers understand pictures and videos, like how we see and recognize things around us.

Data Flow - 5 Stages

1Input Image

1 image x 256 x 256 pixels x 3 color channels→Load and resize image to fixed size→1 image x 256 x 256 pixels x 3 color channels

A photo of a cat resized to 256x256 pixels

↓

2Preprocessing

1 image x 256 x 256 x 3→Normalize pixel values from 0-255 to 0-1→1 image x 256 x 256 x 3

Pixel value 128 becomes 0.5019608

↓

3Feature Extraction

1 image x 256 x 256 x 3→Apply convolution filters to detect edges and shapes→1 image x 64 x 64 x 32 feature maps

Edges of cat ears highlighted in feature maps

↓

4Classification Layer

1 image x 64 x 64 x 32→Flatten and feed to dense layers to predict label→1 vector x 10 classes

Output probabilities for classes like cat, dog, car

↓

5Output Prediction

1 vector x 10→Apply softmax to get probability distribution→1 vector x 10 (probabilities sum to 1)

Cat: 0.85, Dog: 0.10, Car: 0.05

Training Trace - Epoch by Epoch

Loss
1.2 |*       
0.9 | *      
0.7 |  *     
0.5 |   *    
0.4 |    *   
    +---------
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.2	0.45	Model starts learning basic features
2	0.9	0.60	Accuracy improves as edges and shapes are recognized
3	0.7	0.72	Model learns more complex patterns
4	0.5	0.82	Good feature extraction and classification
5	0.4	0.88	Model converges with high accuracy

Prediction Trace - 4 Layers

Layer 1: Input Image

Layer 2: Convolution Layer

Layer 3: Flatten and Dense Layers

Layer 4: Softmax Activation

Model Quiz - 3 Questions

What is the main purpose of the convolution layer in computer vision?

A To increase image size

B To convert images to text

C To detect edges and shapes in images

D To remove colors from images

Key Insight

Computer vision models learn to recognize images by first detecting simple features like edges, then combining them to understand complex shapes, and finally predicting what the image shows with probabilities.