Data

Tools

Indexes

Applications

Experiments

Open Source Science

vast

A repository for some common operations for everyone

https://github.com/vastlab/vast

Science Score: 52.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ✓
    Institutional organization owner
    Organization vastlab has institutional domain (vast.uccs.edu)
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (13.5%) to scientific vocabulary
Last synced: 10 months ago · JSON representation ·

Repository

A repository for some common operations for everyone

Basic Info
  • Host: GitHub
  • Owner: Vastlab
  • License: other
  • Language: Python
  • Default Branch: main
  • Size: 9.91 MB
Statistics
  • Stars: 16
  • Watchers: 10
  • Forks: 5
  • Open Issues: 7
  • Releases: 0
Created over 6 years ago · Last pushed about 2 years ago
Metadata Files
Readme License Citation

README.md

Various Algorithms & Software Tools (VAST)

This repository contains some common functionalities used in various works from the members of the Vision And Security Technology (VAST) Lab.

Setup

For users

pip install git+https://github.com/Vastlab/vast.git

NOTE: There is an unadressed issue due to which the above install makes features like FINCH unable to end user. If you intend to use FINCH please follow the for developers instructions below.

For developers

git clone --recurse-submodules https://github.com/Vastlab/vast.git

pip install -e .[dev]

Contents:

Loss Functions

  1. Entropic Openset loss
  2. Objectosphere loss
  3. Center loss
  4. Objecto-center loss (Objectosphere + Center loss)

Network Architectures

  1. LeNet
  2. LeNet++

Openset Algorithms

  1. OpenMax
  2. Multimodal OpenMax
  3. Extreme Value Machine (EVM)

Reimplementation of libMR

This repo contains a torch based reimplementation of the libMR repo It supports GPU based computation that speeds up the processing considerably, but in certain cases the weibull parameter computation may have slight variations. Reimplementation of libMR

Tools

  1. Concatenate multiple torch datasets Useful for openset learning.
  2. Feature Extraction to HDF5 file from a specific layer of a pytorch model
  3. Multiprocessing Logger

Visualization

  1. 2D visualization e.g. features from LeNet++
  2. 3D visualization for decision planes
  3. OSRC curve plotter
  4. Histogram of scores

Evaluation

  1. OSRC curve using torch tensors and cuda operations
  2. FPR vs coverage plot
  3. Precision/Recall for binary class OOD problem
  4. F-ß Score

Unless a module has a separate license, this code is free for non-commercial use under a BSD-3 style license: We only ask that you cite one our papers, whichever paper is most appropriate from the many.

Examples

Some research works using this repo are:

Self-Supervised Features Improve Open-World Learning

MNIST Based Experiments

ImageNet Level Openset experiments

Owner

  • Name: Vision and Security Technology Lab
  • Login: Vastlab
  • Kind: organization

Citation (CITATION.cff) 

cff-version: 0.0.1
message: "If you use this software, please cite it as below."
authors:
- family-names: "Dhamija"
  given-names: "Akshay"
title: "Various Algorithms & Software Tools (VAST)"
version: 0.0.1
date-released: 2021-10-19
url: "https://github.com/Vastlab/vast"

GitHub Events

Total
  • Watch event: 3
Last Year
  • Watch event: 3

Dependencies

dev-requirements.txt pypi
  • black * development
  • flake8 * development
  • pre-commit * development
requirements.txt pypi
  • glob2 *
  • h5py *
  • matplotlib *
  • numpy *
  • termcolor *
setup.py pypi