Resizing images in Computer Vision - Model Pipeline Trace

Model Pipeline - Resizing images

This pipeline shows how images are resized to a fixed size before being used in a machine learning model. Resizing helps the model handle images of different sizes by making them uniform.

Data Flow - 3 Stages

1Input images

1000 images x varying sizes (e.g., 300x400, 500x600 pixels)→Raw images loaded from dataset→1000 images x varying sizes

Image 1: 300x400 pixels, Image 2: 500x600 pixels

↓

2Resize images

1000 images x varying sizes→Resize each image to 128x128 pixels using bilinear interpolation→1000 images x 128x128 pixels

Image 1 resized from 300x400 to 128x128 pixels

↓

3Normalize pixel values

1000 images x 128x128 pixels→Scale pixel values from 0-255 to 0-1 range→1000 images x 128x128 pixels with normalized values

Pixel value 128 scaled to 0.502

Training Trace - Epoch by Epoch

Loss
1.0 | *       
0.8 |  *      
0.6 |   *     
0.4 |    *    
0.2 |     *   
0.0 +---------
      1 2 3 4 5
       Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.85	0.55	Model starts learning with moderate loss and accuracy
2	0.65	0.7	Loss decreases and accuracy improves as model learns
3	0.5	0.8	Model shows good learning progress
4	0.4	0.85	Loss continues to decrease, accuracy increases
5	0.35	0.88	Model converges with low loss and high accuracy

Prediction Trace - 4 Layers

Layer 1: Input image

Layer 2: Resize operation

Layer 3: Normalization

Layer 4: Model prediction

Model Quiz - 3 Questions

Why do we resize images before feeding them to a model?

A To change the image colors

B To make all images the same size for consistent input

C To increase the image file size

D To remove important details

Key Insight

Resizing images to a fixed size is essential for models to process inputs consistently. Normalizing pixel values helps the model learn better. During training, a good model shows decreasing loss and increasing accuracy, indicating it is learning from the resized images.