Image Classification using Logistic Regression in PyTorch
Part 3 of "PyTorch: Zero to GANs"
This post is the third in a series of tutorials on building deep learning models with PyTorch, an open source neural networks library. Check out the full series:
- PyTorch Basics: Tensors & Gradients
- Linear Regression & Gradient Descent
- Image Classfication using Logistic Regression
- Training Deep Neural Networks on a GPU
- Image Classification using Convolutional Neural Networks
- Data Augmentation, Regularization and ResNets
- Generating Images using Generative Adverserial Networks
In this tutorial, we'll use our existing knowledge of PyTorch and linear regression to solve a very different kind of problem: image classification. We'll use the famous MNIST Handwritten Digits Database as our training dataset. It consists of 28px by 28px grayscale images of handwritten digits (0 to 9), along with labels for each image indicating which digit it represents. Here are some sample images from the dataset:
System setup
This tutorial takes a code-first approach towards learning PyTorch, and you should try to follow along by running and experimenting with the code yourself. The easiest way to start executing this notebook is to click the "Run" button at the top of this page, and select "Run on Kaggle". This will run the notebook on Kaggle, a free online service for running Jupyter notebooks (you might need to create an account).
Running on your computer locally
(Skip this if you're running on Kaggle) To run this notebook locally, clone this notebook, install the required dependencies using conda, and start Jupyter by running the following commands on the terminal / Conda prompt:
pip install jovian --upgrade # Install the jovian library jovian clone aakashns/03-logistic-regression # Download notebook & dependencies cd 03-logistic-regression # Enter the created directory conda create -n 03-logistic-regression python=3.8 # Create an environment conda activate 03-logistic-regression # Activate virtual env jupyter notebook # Start Jupyter
You can find the notebook_id
by cliking the Clone button at the top of this page on Jovian. For a more detailed explanation of the above steps, check out the System setup section in the first notebook.
Exploring the Data
We begin by importing torch
and torchvision
. torchvision
contains some utilities for working with image data. It also contains helper classes to automatically download and import popular datasets like MNIST.
# Uncomment and run the commands below if imports fail
!conda install numpy pytorch torchvision cpuonly -c pytorch -y
!pip install matplotlib --upgrade --quiet
Collecting package metadata (current_repodata.json): done
Solving environment: done
==> WARNING: A newer version of conda exists. <==
current version: 4.7.10
latest version: 4.9.0
Please update conda by running
$ conda update -n base conda
## Package Plan ##
environment location: /srv/conda/envs/notebook
added / updated specs:
- cpuonly
- numpy
- pytorch
- torchvision
The following packages will be downloaded:
package | build
---------------------------|-----------------
blas-1.0 | mkl 6 KB defaults
ca-certificates-2020.6.20 | hecda079_0 145 KB conda-forge
certifi-2020.6.20 | py37he5f6b98_2 151 KB conda-forge
cpuonly-1.0 | 0 2 KB pytorch
freetype-2.10.4 | he06d7ca_0 919 KB conda-forge
intel-openmp-2020.2 | 254 786 KB defaults
jpeg-9d | h516909a_0 266 KB conda-forge
libblas-3.9.0 |1_h6e990d7_netlib 176 KB conda-forge
libcblas-3.9.0 |1_h6e990d7_netlib 51 KB conda-forge
libgfortran-ng-7.5.0 | hae1eefd_17 22 KB conda-forge
libgfortran4-7.5.0 | hae1eefd_17 1.3 MB conda-forge
liblapack-3.9.0 |1_h6e990d7_netlib 2.9 MB conda-forge
libpng-1.6.37 | hed695b0_2 359 KB conda-forge
libtiff-4.0.9 | h648cc4a_1002 566 KB conda-forge
mkl-2020.2 | 256 138.3 MB defaults
ninja-1.10.1 | hfc4b9b4_2 1.9 MB conda-forge
numpy-1.19.2 | py37h7008fea_1 5.2 MB conda-forge
olefile-0.46 | pyh9f0ad1d_1 32 KB conda-forge
openssl-1.1.1h | h516909a_0 2.1 MB conda-forge
pillow-5.4.1 |py37h00a061d_1000 605 KB conda-forge
python_abi-3.7 | 1_cp37m 4 KB conda-forge
pytorch-1.6.0 | py3.7_cpu_0 59.4 MB pytorch
torchvision-0.7.0 | py37_cpu 10.3 MB pytorch
------------------------------------------------------------
Total: 225.3 MB
The following NEW packages will be INSTALLED:
blas pkgs/main/linux-64::blas-1.0-mkl
cpuonly pytorch/noarch::cpuonly-1.0-0
freetype conda-forge/linux-64::freetype-2.10.4-he06d7ca_0
intel-openmp pkgs/main/linux-64::intel-openmp-2020.2-254
jpeg conda-forge/linux-64::jpeg-9d-h516909a_0
libblas conda-forge/linux-64::libblas-3.9.0-1_h6e990d7_netlib
libcblas conda-forge/linux-64::libcblas-3.9.0-1_h6e990d7_netlib
libgfortran-ng conda-forge/linux-64::libgfortran-ng-7.5.0-hae1eefd_17
libgfortran4 conda-forge/linux-64::libgfortran4-7.5.0-hae1eefd_17
liblapack conda-forge/linux-64::liblapack-3.9.0-1_h6e990d7_netlib
libpng conda-forge/linux-64::libpng-1.6.37-hed695b0_2
libtiff conda-forge/linux-64::libtiff-4.0.9-h648cc4a_1002
mkl pkgs/main/linux-64::mkl-2020.2-256
ninja conda-forge/linux-64::ninja-1.10.1-hfc4b9b4_2
numpy conda-forge/linux-64::numpy-1.19.2-py37h7008fea_1
olefile conda-forge/noarch::olefile-0.46-pyh9f0ad1d_1
pillow conda-forge/linux-64::pillow-5.4.1-py37h00a061d_1000
python_abi conda-forge/linux-64::python_abi-3.7-1_cp37m
pytorch pytorch/linux-64::pytorch-1.6.0-py3.7_cpu_0
torchvision pytorch/linux-64::torchvision-0.7.0-py37_cpu
The following packages will be UPDATED:
ca-certificates 2019.6.16-hecc5488_0 --> 2020.6.20-hecda079_0
certifi 2019.6.16-py37_1 --> 2020.6.20-py37he5f6b98_2
openssl 1.1.1c-h516909a_0 --> 1.1.1h-h516909a_0
Downloading and Extracting Packages
ca-certificates-2020 | 145 KB | ##################################### | 100%
freetype-2.10.4 | 919 KB | ##################################### | 100%
liblapack-3.9.0 | 2.9 MB | ##################################### | 100%
jpeg-9d | 266 KB | ##################################### | 100%
blas-1.0 | 6 KB | ##################################### | 100%
libcblas-3.9.0 | 51 KB | ##################################### | 100%
intel-openmp-2020.2 | 786 KB | ##################################### | 100%
ninja-1.10.1 | 1.9 MB | ##################################### | 100%
cpuonly-1.0 | 2 KB | ##################################### | 100%
mkl-2020.2 | 138.3 MB | ##################################### | 100%
certifi-2020.6.20 | 151 KB | ##################################### | 100%
libtiff-4.0.9 | 566 KB | ##################################### | 100%
python_abi-3.7 | 4 KB | ##################################### | 100%
olefile-0.46 | 32 KB | ##################################### | 100%
pytorch-1.6.0 | 59.4 MB | ##################################### | 100%
libgfortran-ng-7.5.0 | 22 KB | ##################################### | 100%
openssl-1.1.1h | 2.1 MB | ##################################### | 100%
pillow-5.4.1 | 605 KB | ##################################### | 100%
libgfortran4-7.5.0 | 1.3 MB | ##################################### | 100%
numpy-1.19.2 | 5.2 MB | ##################################### | 100%
libblas-3.9.0 | 176 KB | ##################################### | 100%
torchvision-0.7.0 | 10.3 MB | ##################################### | 100%
libpng-1.6.37 | 359 KB | ##################################### | 100%
Preparing transaction: done
Verifying transaction: done
Executing transaction: done