Monk - A computer vision toolkit for everyone

Why use Monk

• Issue: Want to begin learning computer vision

  • Solution: Start with Monk's hands-on study roadmap tutorials

• Issue: Multiple libraries hence multiple syntaxes to learn

  • Solution: Monk's one syntax to rule them all - pytorch, keras, mxnet, etc

• Issue: Tough to keep track of all the trial projects while participating in a deep learning competition

  • Solution: Use monk's project management and work on multiple prototyping experiments

• Issue: Tough to set hyper-parameters while training a classifier

  • Solution: Try out hyper-parameter analyser to find the right fit

• Issue: Looking for a library to build quick solutions for your customer

  • Solution: Train, Infer and deploy with monk's low-code syntax

Create real-world Image Classification applications

Medical Domain	Fashion Domain	Autonomous Vehicles Domain
Agriculture Domain	Wildlife Domain	Retail Domain
Satellite Domain	Healthcare Domain	Activity Analysis Domain

...... For more check out the Application Model Zoo!!!!

How does Monk make image classification easy

• Write less code and create end to end applications.
• Learn only one syntax and create applications using any deep learning library - pytorch, mxnet, keras, tensorflow, etc
• Manage your entire project easily with multiple experiments

For whom this library is built

• Students
  • Seamlessly learn computer vision using our comprehensive study roadmaps
• Researchers and Developers
  • Create and Manage multiple deep learning projects
• Competiton participants (Kaggle, Codalab, Hackerearth, AiCrowd, etc)
  • Expedite the prototyping process and jumpstart with a higher rank

Table of Contents

• Sample Showcase
• Installation
• Study Roadmaps, Examples, and Tutorials
• Documentation
• TODO

Sample Showcase - Quick Mode

Create an image classifier.

#Create an experiment
ptf.Prototype("sample-project-1", "sample-experiment-1")

#Load Data
ptf.Default(dataset_path="sample_dataset/", 
             model_name="resnet18", 
             num_epochs=2)
# Train
ptf.Train()

Inference

predictions = ptf.Infer(img_name="sample.png", return_raw=True);

Compare Experiments

#Create comparison project
ctf.Comparison("Sample-Comparison-1");

#Add all your experiments
ctf.Add_Experiment("sample-project-1", "sample-experiment-1");
ctf.Add_Experiment("sample-project-1", "sample-experiment-2");
   
# Generate statistics
ctf.Generate_Statistics();

Installation

• CUDA 9.0          : pip install -U monk-cuda90
• CUDA 9.0          : pip install -U monk-cuda92
• CUDA 10.0        : pip install -U monk-cuda100
• CUDA 10.1        : pip install -U monk-cuda101
• CUDA 10.2        : pip install -U monk-cuda102
• CPU (+Mac-OS) : pip install -U monk-cpu
• Google Colab   : pip install -U monk-colab
• Kaggle              : pip install -U monk-kaggle

For More Installation instructions visit: Link

Study Roadmaps

• Getting started with Monk
  • Essential notebooks to use all the monk's features
• Image Processing and Deep Learning
  • Learn both the basic and advanced concepts of image processing and deep learning
• Transfer Learning
  • Understand transfer learning in the AI field
• Image classification zoo
  • A list of 50+ real world image classification examples

Documentation

• List of available models, layers, blocks, optimizers

• Functional Documentation (Will be merged with Latest docs soon)

  • Main Prototype Functions
    • Mxnet Backend
    • Pytorch Backend
    • Keras Backend
    • Comparison
  • System Functions
  • Mxnet Backend base Functions
  • Pytorch Backend base Functions
  • Keras Backend base Funtions

• Features and Functions (In development):

  • Introduction
  • Setup
  • Quick Mode
  • Update Mode
  • Expert Mode
  • Hyper Parameter Analyser
  • Compare Experiments
  • Resume Training

• Complete Latest Docs (In Progress)

TODO-2020

Features

• Model Visualization
• Pre-processed data visualization
• Learned feature visualization
• NDimensional data input - npy - hdf5 - dicom - tiff
• Multi-label Image Classification
• Custom model development

General

• Functional Documentation
• Tackle Multiple versions of libraries
• Add unit-testing
• Contribution guidelines
• Python pip packaging support

Backend Support

• Tensorflow 2.0 provision support with v1
• Tensorflow 2.0 complete
• Chainer

External Libraries

• TensorRT Acceleration
• Intel Acceleration
• Echo AI - for Activation functions

Connect with the project contributors

Copyright

Copyright 2019 onwards, Tessellate Imaging Private Limited Licensed under the Apache License, Version 2.0 (the "License"); you may not use this project's files except in compliance with the License. A copy of the License is provided in the LICENSE file in this repository.