Monk - A computer vision toolkit for everyone

Monk Object Detection - A low code wrapper over state-of-the-art deep learning algorithms

Why use Monk

• Issue: Abudance of algorithms and difficult to find a working code

  • Solution: All your state-of-the-art as well as old algorithms in one place

• Issue: Installaing different deep learning pipelines is an error-prone task

  • Solution: Single line installations with monk

• Issue: Setting up different algorithms for your custom data requires a lot of effort in changing the existing codes

  • Solution: Easily ingest your custom data for training in COCO, VOC, or Yolo formats

• Issue: Difficulty to trace out which hyperparameters to change for tuning the algorithm

  • Solution: Set your hyper-parameters with a common structure for every algorithm

• Issue: Deployment requires knowledge of base libraries and codes

  • Solution: Easily deploy your models using Monk's low code-syntax

• Issue: Looking for hands-on tutorials for computer vision

  • Solution: Use monk's application building tutorial set

Create real-world Object Detection applications

Wheat detection in field

Detection in underwater imagery

Trash Detection

Object detection in bad lighting

Tiger detection in wild

Person detection in infrared imagery

For more such tutorials visit Application Model Zoo

Create real-world Image Segmentation applications

Road Segmentation in satellite imagery

Ultrasound nerve segmentation

For more such tutorials visit Application Model Zoo

Other applications

Face Detection

Pose Estimation

Activity Recognition

Object Re-identification

Scene Text Localization

Object Tracking

For more such tutorials visit Application Model Zoo

Important Elements

• A) Training Engine

  • Train models on custom dataset witjh low code syntax
  • Pretrained examples on variety of datasets
  • Useful to train your own detector

• B) Inference Engine

  • Original pretrained models (from original authors and implementations) for inferencing and analysing
  • Pretrained models on coco, voc, cityscpaes, type datasets
  • Useful to analyse which algorithm works best for you
  • Useful to generate semi-accurate annotations (coco, pascal-voc, yolo formats) on a new dataset

Training Engine Algorithms

- Train models on custom dataset with low code syntax
- Pretrained examples on variety of datasets
- Useful to train your own detector

NOTE - See the licence file mentioned in the pipelines before using them

Inference Engine Algorithms

- Infer already trained models on COCO/VOC/Open-Images on your custom data
- Useful to analyse computation time metrics

Aknowledgements

• Contributors' information can be found here: https://github.com/Tessellate-Imaging/Monk_Object_Detection/blob/master/Contributors.md
• Majority of the code is obtained from these pipelines (Monk is a low code wrapper to utilize these pipelines)
  • https://gluon-cv.mxnet.io/build/examples_detection/index.html
  • https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf1.md
  • https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2.md
  • https://github.com/signatrix/efficientdet
  • https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch
  • https://pytorch.org/tutorials/intermediate/torchvision_tutorial.html
  • https://github.com/ijkguo/mx-rcnn
  • https://github.com/yhenon/pytorch-retinanet
  • https://github.com/princeton-vl/CornerNet-Lite
  • https://github.com/ultralytics/yolov3
  • https://github.com/ruinmessi/RFBNet
  • https://github.com/PengyiZhang/SlimYOLOv3
  • https://github.com/qfgaohao/pytorch-ssd
  • https://github.com/open-mmlab/mmdetection
  • https://github.com/qubvel/segmentation_models
  • https://github.com/biubug6/Pytorch_Retinaface
  • https://github.com/open-mmlab/mmaction2
  • https://github.com/WXinlong/SOLO
  • https://gluon-cv.mxnet.io/build/examples_pose/dive_deep_simple_pose.html
  • https://github.com/Cartucho/mAP

Author

Tessellate Imaging - https://www.tessellateimaging.com/

Check out Monk AI - (https://github.com/Tessellate-Imaging/monk_v1)

Monk features
    - low-code
    - unified wrapper over major deep learning framework - keras, pytorch, gluoncv
    - syntax invariant wrapper

Enables developers
    - to create, manage and version control deep learning experiments
    - to compare experiments across training metrics
    - to quickly find best hyper-parameters

To contribute to Monk AI or Monk Object Detection repository raise an issue in the git-repo or dm us on linkedin

• Abhishek - https://www.linkedin.com/in/abhishek-kumar-annamraju/
• Akash - https://www.linkedin.com/in/akashdeepsingh01/

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.