Model comparison in Computer Vision - Model Pipeline Trace

Model Pipeline - Model comparison

This pipeline compares two image classification models to see which one performs better on the same task. It shows how data flows, how training improves each model, and how predictions differ.

Data Flow - 6 Stages

1Input images

1000 images x 64 x 64 pixels x 3 channels→Raw images loaded for classification→1000 images x 64 x 64 pixels x 3 channels

Image of a cat with RGB colors

↓

2Preprocessing

1000 images x 64 x 64 x 3→Normalize pixel values to 0-1 range→1000 images x 64 x 64 x 3

Pixel value 128 becomes 0.5

↓

3Feature extraction Model A

1000 images x 64 x 64 x 3→Convolutional layers extract features→1000 images x 16 x 16 x 32

Edges and shapes detected in images

↓

4Feature extraction Model B

1000 images x 64 x 64 x 3→Deeper convolutional layers extract features→1000 images x 8 x 8 x 64

More detailed features detected

↓

5Flatten and Dense layers Model A

1000 images x 16 x 16 x 32→Flatten features and classify→1000 images x 10 classes

Output probabilities for 10 classes

↓

6Flatten and Dense layers Model B

1000 images x 8 x 8 x 64→Flatten features and classify→1000 images x 10 classes

Output probabilities for 10 classes

Training Trace - Epoch by Epoch

Epochs: 1  2  3
Model A Loss: 1.2-0.9-0.7
Model B Loss: 1.0-0.6-0.4
Loss decreases steadily for both models, Model B faster.

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.2	0.55	Model A starts with moderate accuracy and high loss
2	0.9	0.68	Model A improves as it learns features
3	0.7	0.75	Model A continues to improve steadily
1	1.0	0.60	Model B starts slightly better than Model A
2	0.6	0.78	Model B learns faster with deeper layers
3	0.4	0.85	Model B shows stronger performance

Prediction Trace - 5 Layers

Layer 1: Input image

Layer 2: Feature extraction Model A

Layer 3: Classification Model A

Layer 4: Feature extraction Model B

Layer 5: Classification Model B

Model Quiz - 3 Questions

Which model shows faster improvement in accuracy during training?

A Both improve equally

B Model A

C Model B

D Neither improves

Key Insight

Comparing two models side-by-side helps us see which learns faster and predicts better. Model B, with deeper layers, improves accuracy quicker and predicts with higher confidence, showing the value of more complex feature extraction in image tasks.