Cropping images in Computer Vision - Model Pipeline Trace

Model Pipeline - Cropping images

This pipeline shows how images are cropped to focus on important parts before being used in a machine learning model. Cropping helps the model learn better by removing unnecessary background.

Data Flow - 4 Stages

1Input images

1000 images x 256 height x 256 width x 3 channels→Load raw images of size 256x256 pixels with 3 color channels (RGB)→1000 images x 256 height x 256 width x 3 channels

An image of a cat with full background

↓

2Cropping

1000 images x 256 height x 256 width x 3 channels→Crop center 128x128 pixels from each image to focus on main object→1000 images x 128 height x 128 width x 3 channels

Cropped image showing only the cat's face

↓

3Normalization

1000 images x 128 height x 128 width x 3 channels→Scale pixel values from 0-255 to 0-1 range→1000 images x 128 height x 128 width x 3 channels

Pixel value 128 becomes 0.502

↓

4Model input

1000 images x 128 height x 128 width x 3 channels→Feed cropped and normalized images into the model→1000 predictions x number_of_classes

Model predicts class probabilities for each image

Training Trace - Epoch by Epoch

Loss
1.2 |****
1.0 |*** 
0.8 |**  
0.6 |*   
0.4 |*   
    +-----
     1 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.2	0.45	Model starts learning with moderate loss and low accuracy
2	0.9	0.60	Loss decreases and accuracy improves as model learns features
3	0.7	0.72	Model continues to improve with better focus on cropped images
4	0.5	0.80	Loss drops further and accuracy reaches a good level
5	0.4	0.85	Training converges with low loss and high accuracy

Prediction Trace - 4 Layers

Layer 1: Input image

Layer 2: Cropping

Layer 3: Normalization

Layer 4: Model prediction

Model Quiz - 3 Questions

Why do we crop images before training the model?

A To add noise for data augmentation

B To increase image size for better detail

C To focus on important parts and remove background

D To convert images to grayscale

Key Insight

Cropping images helps the model focus on the main object by removing unnecessary background. This leads to better learning and higher accuracy as seen by the decreasing loss and increasing accuracy during training.