Compose transforms in PyTorch - Model Pipeline Trace

Model Pipeline - Compose transforms

This pipeline shows how multiple image transformations are combined using Compose in PyTorch to prepare data for training a model. Compose applies each transform step-by-step to the input images, making preprocessing easy and organized.

Data Flow - 3 Stages

1Raw Image Input

1000 images x 3 channels x 256 height x 256 width→Original dataset of color images→1000 images x 3 channels x 256 height x 256 width

An image of a cat with size 256x256 pixels and 3 color channels (RGB)

↓

2Compose Transforms Applied

1000 images x 3 channels x 256 height x 256 width→Resize to 128x128, convert to tensor, normalize pixel values→1000 images x 3 channels x 128 height x 128 width

The cat image resized smaller, pixel values scaled to range [-1,1]

↓

3Batch Formation

1000 images x 3 channels x 128 height x 128 width→Group images into batches of 32 for training→32 images x 3 channels x 128 height x 128 width

A batch containing 32 resized and normalized cat images

Training Trace - Epoch by Epoch

Epochs
1 |***************
2 |************
3 |*********
4 |******
5 |****
Loss

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.2	0.45	Initial training with high loss and low accuracy
2	0.9	0.60	Loss decreased, accuracy improved as model learns
3	0.7	0.72	Continued improvement in training metrics
4	0.55	0.80	Model is converging with better accuracy
5	0.45	0.85	Training loss low and accuracy high, good convergence

Prediction Trace - 4 Layers

Layer 1: Resize

Layer 2: ToTensor

Layer 3: Normalize

Layer 4: Model Prediction

Model Quiz - 3 Questions

What does the Compose transform do in this pipeline?

A Trains the model on batches

B Splits data into training and testing sets

C Applies multiple image transformations in sequence

D Calculates accuracy after each epoch

Key Insight

Using Compose to chain transforms simplifies preprocessing and ensures consistent input format. This helps the model learn better, shown by decreasing loss and increasing accuracy during training.