Skip to content

Commit

Permalink
update image formats
Browse files Browse the repository at this point in the history
  • Loading branch information
@dcarpintero
dcarpintero committed Apr 11, 2024
1 parent 853ae57 commit 0b91dad
Showing 1 changed file with 56 additions and 46 deletions.
102 changes: 56 additions & 46 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,31 +17,34 @@ In this article, we will implement in Python the essential modules required to b

By the end of this guide, you will be able to construct the building blocks of a neural network from scratch, understand how it learns, and deploy it to [HuggingFace Spaces](https://huggingface.co/spaces/dcarpintero/fashion-image-recognitionhttps://huggingface.co/spaces/dcarpintero/fashion-image-recognition) to classify real-world garment images.

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/bvC2A3Cb2zn81h_neojtH.png">
<figcaption style="text-align: center;">Garment Classifier deployed to HuggingFace Spaces</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/bvC2A3Cb2zn81h_neojtH.png">
</p>

<p align="center">Garment Classifier deployed to HuggingFace Spaces</p>

## Table of Contents

- [The Intuition behind our Neural Network](#the-intuition-behind-our-neuralnetwork)
- [Architecture](#architecture)
- [Linear Transformation](#linear-transformation)
- [Introducing non-linearity](#introducing-non-linearity)
- [Regularization](#regularization)
- [Flatten Transformation](#flatten-transformation)
- [Sequential Layer](#sequential-layer)
- [Classifier Model](#classifier-model)
- [Gradient Descent Optimizer](#gradient-descent-optimizer)
- [Backpropagation](#backpropagation)
- [Training](#training)
- [The Fashion Dataset](#the-fashion-dataset)
- [Data Loaders for Mini-Batches](#data-loaders-for-mini-batches)
- [Fitting the Model](#fitting-the-model)
- [Model Assessment](#model-assessment)
- [Inference](#inference)
- [Resources](#resources)
- [References](#references)
- [Building a Neural Network Classifier from the Ground Up: A Step-by-Step Guide](#building-a-neural-network-classifier-from-the-ground-up-a-step-by-step-guide)
- [Table of Contents](#table-of-contents)
- [The Intuition behind our Neural Network](#the-intuition-behind-our-neuralnetwork)
- [Architecture](#architecture)
- [Linear Transformation](#linear-transformation)
- [Introducing non-linearity](#introducing-non-linearity)
- [Regularization](#regularization)
- [Flatten Transformation](#flatten-transformation)
- [Sequential Layer](#sequential-layer)
- [Classifier Model](#classifier-model)
- [Gradient Descent Optimizer](#gradient-descent-optimizer)
- [Backpropagation](#backpropagation)
- [Training](#training)
- [The Fashion Dataset](#the-fashion-dataset)
- [Data Loaders for Mini-Batches](#data-loaders-for-mini-batches)
- [Fitting the Model](#fitting-the-model)
- [Model Assessment](#model-assessment)
- [Inference](#inference)
- [Resources](#resources)
- [References](#references)



Expand Down Expand Up @@ -241,10 +244,12 @@ class Classifier(nn.Module):

The research paper *Visualizing and Understanding Convolutional Networks* [7] offers insights into a concept akin to hierarchical progressive learning, specifically applied to convolutional layers. This provides a comparable intuition to understand how stacked layers are capable of automatically learning features within images:

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/CMgQHSSzEaEmsBd0D2eKD.png">
<figcaption style="text-align: center;">Visualization of features in a convolutional neural network - https://arxiv.org/pdf/1311.2901.pdf</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/CMgQHSSzEaEmsBd0D2eKD.png">
</p>

<p align="center">Visualization of features in a convolutional neural network - https://arxiv.org/pdf/1311.2901.pdf</p>


### Gradient Descent Optimizer

Expand Down Expand Up @@ -406,31 +411,35 @@ class Learner:

After 25 epochs, our model achieves 0.868 accuracy, which fairly approximates [benchmark results](http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/#) (0.874 for an MLP Classifier using ReLU as the activation function).

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/BMTJGJabSWi3Jcoc_m5sH.png">
<figcaption style="text-align: center;">Model Assessment (epochs=25, lr=0.005, batch_size=32, SGD, CrossEntropyLoss) w/ self-implemented modules</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/BMTJGJabSWi3Jcoc_m5sH.png" width=700>
</p>

<p align="center">Model Assessment (epochs=25, lr=0.005, batch_size=32, SGD, CrossEntropyLoss) w/ self-implemented modules</p>

We observe comparable accuracy levels between our self-implemented modules and a standard PyTorch implementation with the same hyperparameters (`epochs=25, lr=0.005, batch_size=32`). Notably, the PyTorch model demonstrates a slighly smaller gap between validation and training losses, suggesting better generalization capabilities:

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/mCLHZnCYArxC8YcSdEjYA.png">
<figcaption style="text-align: center;">Model Assessment (epochs=25, lr=0.005, batch_size=32, SGD, CrossEntropyLoss) w/ PyTorch modules</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/mCLHZnCYArxC8YcSdEjYA.png" width=700>
</p>

<p align="center">Model Assessment (epochs=25, lr=0.005, batch_size=32, SGD, CrossEntropyLoss) w/ PyTorch modules</p>

Furthermore, a basic analysis of precision (accuracy of the positive predictions for a specific class), recall (ability to detect all relevant instances of a specific class), and f1-score (mean of precision and recall) reveals that our model excels in categories with distinctive features such as Trouser/Jeans, Sandal, Bag, and Ankle-Boot. However, it performs below average with Shirts, Pullovers, and Coats.

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/amnSh3QHhkgWtRKzQUpVu.png">
<figcaption style="text-align: center;">Precision, Recall and F1-Scores across Categories</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/amnSh3QHhkgWtRKzQUpVu.png" width=400>
</p>

<p align="center">Precision, Recall and F1-Scores across Categories</p>

The confussion matrix confirms that the Shirt category is frequently confused with the T-Shirt/Top, Pullover, and Coat classes; whereas Coat is confused with Shirt and Pullover. This suggests that working at 28x28 pixels resolution might cast upper body garment categories as visually challenging.

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/KqbEVgVv8gz5X8JkZzX4H.png">
<figcaption style="text-align: center;">Confussion Matrix</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/KqbEVgVv8gz5X8JkZzX4H.png" width=650>
</p>

<p align="center">Confussion Matrix</p>

## Inference

Expand All @@ -456,10 +465,11 @@ transform = transforms.Compose(

This can be easily integrated into a Gradio App, and then deployed to [HuggingFace Spaces](https://huggingface.co/spaces/dcarpintero/fashion-image-recognition):

<figure>
<img style="margin: 0 auto; display: block;" src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/HwZwQbtOuuJx-VrhnV-YD.png">
<figcaption style="text-align: center;">Garment Classifier deployed to HuggingFace Spaces</figcaption>
</figure>
<p align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a13b68b14ab77f9e3eb061/HwZwQbtOuuJx-VrhnV-YD.png" width=650>
</p>

<p align="center">Garment Classifier deployed to HuggingFace Spaces</p>

## Resources

Expand Down

0 comments on commit 0b91dad

Please sign in to comment.