Image augmentation transforms in Computer Vision - Model Pipeline Trace

Model Pipeline - Image augmentation transforms

This pipeline shows how image data is changed using augmentation transforms to help a model learn better. Augmentation creates new images by flipping, rotating, or changing brightness, so the model sees more variety and becomes stronger.

Data Flow - 5 Stages

1Original Images

1000 images x 64 x 64 x 3→Raw image data loaded from dataset→1000 images x 64 x 64 x 3

An image of a cat with size 64x64 pixels and 3 color channels (RGB)

↓

2Random Horizontal Flip

1000 images x 64 x 64 x 3→Flip each image horizontally with 50% chance→1000 images x 64 x 64 x 3

Cat image flipped left to right or kept as is

↓

3Random Rotation

1000 images x 64 x 64 x 3→Rotate images randomly between -15 to +15 degrees→1000 images x 64 x 64 x 3

Cat image rotated slightly clockwise or counterclockwise

↓

4Random Brightness Adjustment

1000 images x 64 x 64 x 3→Change brightness randomly by ±20%→1000 images x 64 x 64 x 3

Cat image appears lighter or darker

↓

5Augmented Images

1000 images x 64 x 64 x 3→Combined augmentations applied to increase data variety→1000 images x 64 x 64 x 3

Cat images with flips, rotations, and brightness changes

Training Trace - Epoch by Epoch

Loss
1.2 |****
0.9 |***
0.7 |**
0.55|*
0.45| 
    +------------
    Epochs 1 to 5

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.2	0.45	Model starts learning with high loss and low accuracy
2	0.9	0.60	Loss decreases and accuracy improves as model learns
3	0.7	0.72	Model continues to improve with augmented data
4	0.55	0.80	Augmentation helps model generalize better
5	0.45	0.85	Loss lowers steadily and accuracy reaches good level

Prediction Trace - 5 Layers

Layer 1: Input Image

Layer 2: Random Horizontal Flip

Layer 3: Random Rotation

Layer 4: Random Brightness Adjustment

Layer 5: Model Input

Model Quiz - 3 Questions

What is the main purpose of applying image augmentation transforms?

A To make images black and white

B To increase the variety of training images

C To reduce the size of the dataset

D To remove noise from images

Key Insight

Image augmentation helps the model see many different versions of the same image. This variety teaches the model to recognize objects better in new, unseen pictures, improving accuracy and reducing overfitting.