amritpurshotam / computer-vision-papers Goto Github PK

View Code? Open in Web Editor NEW

0.0 0.0 0.0 1.33 MB

Reproduction of the architectures and results of the most significant computer vision papers

License: GNU General Public License v3.0

Python 1.68% Jupyter Notebook 98.32%

computer-vision deep-learning python tensorflow

computer-vision-papers's Introduction

👋 Hi there

My name is Amrit Purshotam and I'm a machine learning scientist.

🛠️ What I'm building

Here are some projects that I work on in my spare time.

ImageNet Papers: I'm implementing the best performing architectures from every year of the ImageNet challenge since 2012 as an exercise in reproduction and to step through time to see how the approaches evolved first-hand. What I'm doing differently is implementing the architectures with the full training pipelines and making the training runs public with wandb.
FancyPCA: I built a demo of the PCA Colour Augmentation technique with Streamlit which can be seen live here.
Baobab: I'm helping to build an event management platform that enables grass-roots community builders to scale their efforts when organising conferences that have complex requirements for selecting attendees. The application is primarily built for and used by the Deep Learning Indaba who I passionately support and helping build this is my way of contributing to their mission.

I'm also currently incubating the below projects while I figure out some final details before getting started in earnest.

DeepSARA: A revisit of my MSc project where I built classification and detection models for southern African rock art figures using classical ML and CV techniques. I'm now curious to see how well some deep learning approaches work. But first I need to re-label the dataset.
park2vec: I want to train general location vectors with the aim to use them to build a parkrun course similarity tool. I'm currently trying to wrangle OpenStreetMap data into a format that would be useful to train on.

📈 GitHub Stats

computer-vision-papers's People

Watchers

computer-vision-papers's Issues

Recalculate mean on ImageNet and also calculate standard deviation

Currently we're resizing and cropping before taking the mean. Remove this and just read in the image then take the mean and standard deviation.

Also for reference the Pytorch mean and standard deviation is as follows

normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                 std=[0.229, 0.224, 0.225])

Add some extra linters

flake8-docstrings (wraps pydocstyle)). Forces you to use comments on every function though.
flake8-broken-line
darglint
flake8-eradicate
vulture
pep8-naming

Nice summary of all the error codes here
https://wemake-python-stylegui.de/en/latest/pages/usage/violations/index.html

Investigate how tf.image.decode_jpeg handles BW, CMYK, RGBA images compared to pillow

Use newer optimizers for training speed up and increased training accuracy on AlexNet

See the top comment for reference.

Standardise images in AlexNet to increase accuracy

See top comment here for reference.

Use better initialisation to improve training accuracy

Unable to install requirements.txt

Hi Amrit, I was trying to install your project to test whether my fixes to mllint worked, but I noticed that it was not possible to run pip install -e . due to unresolvable dependencies. It seems pip is primarily having trouble with GitPython.

Note that I'm using Python 3.9.6 and Pip nowadays has a fancy new dependency resolver that might be struggling here.

See the console output of `pip install -e .` below.

 ~/tudelft/thesis/computer-vision-papers > main > pip install -e .
Defaulting to user installation because normal site-packages is not writeable
Obtaining file:///home/bart/tudelft/thesis/computer-vision-papers
Requirement already satisfied: click==8.0.1 in /usr/lib/python3.9/site-packages (from src==0.1.0) (8.0.1)
Collecting matplotlib==3.4.2
  Using cached matplotlib-3.4.2-cp39-cp39-manylinux1_x86_64.whl (10.3 MB)
Collecting numpy==1.19.2
  Using cached numpy-1.19.2.zip (7.3 MB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
    Preparing wheel metadata ... done
Collecting tensorflow==2.5.0
  Using cached tensorflow-2.5.0-cp39-cp39-manylinux2010_x86_64.whl (454.4 MB)
Collecting tensorflow-datasets==4.4.0
  Using cached tensorflow_datasets-4.4.0-py3-none-any.whl (4.0 MB)
Collecting wandb==0.11.2
  Using cached wandb-0.11.2-py2.py3-none-any.whl (1.8 MB)
Requirement already satisfied: cycler>=0.10 in /home/bart/.local/lib/python3.9/site-packages (from matplotlib==3.4.2->src==0.1.0) (0.10.0)
Requirement already satisfied: python-dateutil>=2.7 in /home/bart/.local/lib/python3.9/site-packages (from matplotlib==3.4.2->src==0.1.0) (2.8.1)
Requirement already satisfied: pyparsing>=2.2.1 in /usr/lib/python3.9/site-packages (from matplotlib==3.4.2->src==0.1.0) (2.4.7)
Requirement already satisfied: kiwisolver>=1.0.1 in /home/bart/.local/lib/python3.9/site-packages (from matplotlib==3.4.2->src==0.1.0) (1.3.1)
Requirement already satisfied: pillow>=6.2.0 in /usr/lib/python3.9/site-packages (from matplotlib==3.4.2->src==0.1.0) (8.3.1)
Collecting absl-py~=0.10
  Using cached absl_py-0.13.0-py3-none-any.whl (132 kB)
Collecting six~=1.15.0
  Using cached six-1.15.0-py2.py3-none-any.whl (10 kB)
Collecting grpcio~=1.34.0
  Using cached grpcio-1.34.1-cp39-cp39-manylinux2014_x86_64.whl (4.0 MB)
Collecting opt-einsum~=3.3.0
  Using cached opt_einsum-3.3.0-py3-none-any.whl (65 kB)
Collecting wheel~=0.35
  Using cached wheel-0.37.0-py2.py3-none-any.whl (35 kB)
Collecting flatbuffers~=1.12.0
  Using cached flatbuffers-1.12-py2.py3-none-any.whl (15 kB)
Collecting tensorboard~=2.5
  Using cached tensorboard-2.6.0-py3-none-any.whl (5.6 MB)
Collecting typing-extensions~=3.7.4
  Using cached typing_extensions-3.7.4.3-py3-none-any.whl (22 kB)
Collecting astunparse~=1.6.3
  Using cached astunparse-1.6.3-py2.py3-none-any.whl (12 kB)
Collecting tensorflow-estimator<2.6.0,>=2.5.0rc0
  Using cached tensorflow_estimator-2.5.0-py2.py3-none-any.whl (462 kB)
Requirement already satisfied: termcolor~=1.1.0 in /usr/lib/python3.9/site-packages (from tensorflow==2.5.0->src==0.1.0) (1.1.0)
Collecting google-pasta~=0.2
  Using cached google_pasta-0.2.0-py3-none-any.whl (57 kB)
Requirement already satisfied: wrapt~=1.12.1 in /home/bart/.local/lib/python3.9/site-packages (from tensorflow==2.5.0->src==0.1.0) (1.12.1)
Collecting gast==0.4.0
  Using cached gast-0.4.0-py3-none-any.whl (9.8 kB)
Collecting h5py~=3.1.0
  Using cached h5py-3.1.0-cp39-cp39-manylinux1_x86_64.whl (4.4 MB)
Collecting protobuf>=3.9.2
  Using cached protobuf-3.17.3-cp39-cp39-manylinux_2_5_x86_64.manylinux1_x86_64.whl (1.0 MB)
Collecting keras-preprocessing~=1.1.2
  Using cached Keras_Preprocessing-1.1.2-py2.py3-none-any.whl (42 kB)
Collecting keras-nightly~=2.5.0.dev
  Using cached keras_nightly-2.5.0.dev2021032900-py2.py3-none-any.whl (1.2 MB)
Requirement already satisfied: attrs>=18.1.0 in /usr/lib/python3.9/site-packages (from tensorflow-datasets==4.4.0->src==0.1.0) (21.2.0)
Requirement already satisfied: requests>=2.19.0 in /usr/lib/python3.9/site-packages (from tensorflow-datasets==4.4.0->src==0.1.0) (2.26.0)
Collecting dill
  Using cached dill-0.3.4-py2.py3-none-any.whl (86 kB)
Collecting promise
  Using cached promise-2.3.tar.gz (19 kB)
Collecting tensorflow-metadata
  Using cached tensorflow_metadata-1.2.0-py3-none-any.whl (48 kB)
Collecting future
  Using cached future-0.18.2-py3-none-any.whl
Collecting tqdm
  Using cached tqdm-4.62.0-py2.py3-none-any.whl (76 kB)
Collecting shortuuid>=0.5.0
  Using cached shortuuid-1.0.1-py3-none-any.whl (7.5 kB)
Collecting pathtools
  Using cached pathtools-0.1.2.tar.gz (11 kB)
Collecting subprocess32>=3.5.3
  Using cached subprocess32-3.5.4.tar.gz (97 kB)
Requirement already satisfied: PyYAML in /usr/lib/python3.9/site-packages (from wandb==0.11.2->src==0.1.0) (5.4.1)
Requirement already satisfied: urllib3>=1.26.5 in /usr/lib/python3.9/site-packages (from wandb==0.11.2->src==0.1.0) (1.26.6)
Collecting sentry-sdk>=1.0.0
  Using cached sentry_sdk-1.3.1-py2.py3-none-any.whl (133 kB)
Requirement already satisfied: GitPython>=1.0.0 in /home/bart/.local/lib/python3.9/site-packages (from wandb==0.11.2->src==0.1.0) (3.1.14)
Collecting configparser>=3.8.1
  Using cached configparser-5.0.2-py3-none-any.whl (19 kB)
Requirement already satisfied: psutil>=5.0.0 in /usr/lib/python3.9/site-packages (from wandb==0.11.2->src==0.1.0) (5.8.0)
Requirement already satisfied: docker-pycreds>=0.4.0 in /usr/lib/python3.9/site-packages (from wandb==0.11.2->src==0.1.0) (0.4.0)
Requirement already satisfied: gitdb<5,>=4.0.1 in /home/bart/.local/lib/python3.9/site-packages (from GitPython>=1.0.0->wandb==0.11.2->src==0.1.0) (4.0.7)
Requirement already satisfied: smmap<5,>=3.0.1 in /home/bart/.local/lib/python3.9/site-packages (from gitdb<5,>=4.0.1->GitPython>=1.0.0->wandb==0.11.2->src==0.1.0) (4.0.0)
INFO: pip is looking at multiple versions of grpcio to determine which version is compatible with other requirements. This could take a while.
Collecting grpcio~=1.34.0
  Using cached grpcio-1.34.0-cp39-cp39-manylinux2014_x86_64.whl (4.0 MB)
INFO: pip is looking at multiple versions of google-pasta to determine which version is compatible with other requirements. This could take a while.
INFO: pip is looking at multiple versions of gitdb to determine which version is compatible with other requirements. This could take a while.
Collecting gitdb<5,>=4.0.1
  Using cached gitdb-4.0.7-py3-none-any.whl (63 kB)
  Using cached gitdb-4.0.6-py3-none-any.whl (63 kB)
Collecting smmap<4,>=3.0.1
  Using cached smmap-3.0.5-py2.py3-none-any.whl (25 kB)
Collecting gitdb<5,>=4.0.1
  Using cached gitdb-4.0.5-py3-none-any.whl (63 kB)
INFO: pip is looking at multiple versions of grpcio to determine which version is compatible with other requirements. This could take a while.
  Using cached gitdb-4.0.4-py3-none-any.whl (63 kB)
  Using cached gitdb-4.0.2-py3-none-any.whl (63 kB)
  Using cached gitdb-4.0.1-py2.py3-none-any.whl (63 kB)
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking
INFO: pip is looking at multiple versions of gitpython to determine which version is compatible with other requirements. This could take a while.
Collecting GitPython>=1.0.0
  Using cached GitPython-3.1.18-py3-none-any.whl (170 kB)
INFO: pip is looking at multiple versions of google-pasta to determine which version is compatible with other requirements. This could take a while.
INFO: pip is looking at multiple versions of gitdb to determine which version is compatible with other requirements. This could take a while.
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking
  Using cached GitPython-3.1.14-py3-none-any.whl (159 kB)
  Using cached GitPython-3.1.13-py3-none-any.whl (159 kB)
  Using cached GitPython-3.1.12-py3-none-any.whl (159 kB)
  Using cached GitPython-3.1.11-py3-none-any.whl (159 kB)
  Using cached GitPython-3.1.10-py3-none-any.whl (159 kB)
  Using cached GitPython-3.1.9-py3-none-any.whl (159 kB)
INFO: pip is looking at multiple versions of gitpython to determine which version is compatible with other requirements. This could take a while.
  Using cached GitPython-3.1.8-py3-none-any.whl (159 kB)
  Using cached GitPython-3.1.7-py3-none-any.whl (158 kB)
  Using cached GitPython-3.1.6-py3-none-any.whl (158 kB)
  Using cached GitPython-3.1.5-py3-none-any.whl (451 kB)
  Using cached GitPython-3.1.3-py3-none-any.whl (451 kB)
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking
  Using cached GitPython-3.1.2-py3-none-any.whl (451 kB)
  Using cached GitPython-3.1.1-py3-none-any.whl (450 kB)
  Using cached GitPython-3.1.0-py3-none-any.whl (450 kB)
  Using cached GitPython-3.0.9-py3-none-any.whl (450 kB)
Collecting gitdb2<4,>=3
  Using cached gitdb2-3.0.3.post1-py2.py3-none-any.whl (63 kB)
Collecting smmap2<3,>=2
  Using cached smmap2-2.0.5-py2.py3-none-any.whl (24 kB)
INFO: pip is looking at multiple versions of gitdb2 to determine which version is compatible with other requirements. This could take a while.
Collecting gitdb2<4,>=3
  Using cached gitdb2-3.0.3-py2.py3-none-any.whl (63 kB)
  Using cached gitdb2-3.0.2-py2.py3-none-any.whl (63 kB)
Collecting smmap2>=2.0.0
  Using cached smmap2-3.0.1-py3-none-any.whl (1.1 kB)
Collecting gitdb2<4,>=3
  Using cached gitdb2-3.0.1-py2.py3-none-any.whl (63 kB)
  Using cached gitdb2-3.0.0-py2.py3-none-any.whl (63 kB)
Collecting GitPython>=1.0.0
  Using cached GitPython-3.0.8-py3-none-any.whl (450 kB)
Collecting gitdb2>=3
  Using cached gitdb2-4.0.2-py3-none-any.whl (1.1 kB)
INFO: pip is looking at multiple versions of gitdb2 to determine which version is compatible with other requirements. This could take a while.
Collecting GitPython>=1.0.0
  Using cached GitPython-3.0.7-py3-none-any.whl (451 kB)
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking
Collecting gitdb2>=2.0.0
  Using cached gitdb2-2.0.6-py2.py3-none-any.whl (63 kB)
  Using cached gitdb2-2.0.5-py2.py3-none-any.whl (62 kB)
  Using cached gitdb2-2.0.4-py2.py3-none-any.whl (62 kB)
  Using cached gitdb2-2.0.3-py2.py3-none-any.whl (63 kB)
  Using cached gitdb2-2.0.2-py2.py3-none-any.whl (63 kB)
  Using cached gitdb2-2.0.1-py2.py3-none-any.whl (63 kB)
  Using cached gitdb2-2.0.0-py2.py3-none-any.whl (63 kB)
Collecting GitPython>=1.0.0
  Using cached GitPython-3.0.6-py3-none-any.whl (451 kB)
  Using cached GitPython-3.0.5-py3-none-any.whl (455 kB)
  Using cached GitPython-3.0.4-py3-none-any.whl (454 kB)
  Using cached GitPython-3.0.3-py3-none-any.whl (453 kB)
  Using cached GitPython-3.0.2-py3-none-any.whl (453 kB)
  Using cached GitPython-3.0.1-py3-none-any.whl (453 kB)
Collecting ddt>=1.1.1
  Using cached ddt-1.4.2-py2.py3-none-any.whl (6.1 kB)
INFO: pip is looking at multiple versions of ddt to determine which version is compatible with other requirements. This could take a while.
  Using cached ddt-1.4.1-py2.py3-none-any.whl (6.1 kB)
  Using cached ddt-1.4.0-py2.py3-none-any.whl (6.1 kB)
  Using cached ddt-1.3.1-py2.py3-none-any.whl (5.6 kB)
  Using cached ddt-1.3.0-py2.py3-none-any.whl (5.6 kB)
  Using cached ddt-1.2.2-py2.py3-none-any.whl (5.4 kB)
  Using cached ddt-1.2.1-py2.py3-none-any.whl (5.4 kB)
  Using cached ddt-1.2.0-py2.py3-none-any.whl (4.3 kB)
INFO: pip is looking at multiple versions of ddt to determine which version is compatible with other requirements. This could take a while.
  Using cached ddt-1.1.3-py2.py3-none-any.whl (4.2 kB)
  Using cached ddt-1.1.2-py2.py3-none-any.whl (5.1 kB)
  Using cached ddt-1.1.1-py2.py3-none-any.whl (5.0 kB)
Collecting GitPython>=1.0.0
  Using cached GitPython-3.0.0-py3-none-any.whl (452 kB)
Collecting gitdb>=0.6.4
  Using cached gitdb-0.6.4.tar.gz (400 kB)
INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking
Collecting GitPython>=1.0.0
  Using cached GitPython-2.1.15-py2.py3-none-any.whl (452 kB)
  Using cached GitPython-2.1.14-py2.py3-none-any.whl (452 kB)
  Using cached GitPython-2.1.13-py2.py3-none-any.whl (452 kB)
  Using cached GitPython-2.1.12-py2.py3-none-any.whl (452 kB)
^CERROR: Operation cancelled by user
WARNING: You are using pip version 21.1.2; however, version 21.2.3 is available.
You should consider upgrading via the '/usr/bin/python -m pip install --upgrade pip' command.

Log gradients with subset of training set

Test whether wandb uses the full training set when logging the gradients. If so and it's very slow (takes approx 1hr to even read all the data from disk), then setup another dataset that only uses a smaller subset (maybe 5000 random images so hopefully at least one image from each class is selected)

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.