Smart Ordering

A smart application to track multiple person and arrange them according to the time of arrival based on efficientdet

[1] Abd al-Rahman al-Ktefane, Adel Kaboul, Ammar Abo Azan, Oday Mourad. Smart Ordering: Scalable and Efficient Person Tracker and Sorter 2020. Paper Link

Quick install dependencies: pip install -r requirements.txt

1. install dependencies

pip install -r requirements.txt

2. Export efficientDet0 SavedModel, frozen graph.

Run the following command line to export models: ps: exclamation point "!" used for run command's in colab cell, if you run it on regular shell please remove it.

!rm  -rf resources/savedmodeldir
!rm  -rf resources/efficientdet-d0
!tar -zxvf resources/efficientdet-d0.tar.gz
!python model_inspect.py --runmode=saved_model --model_name=efficientdet-d0 \
  --ckpt_path=resources/efficientdet-d0 --saved_model_dir=resources/savedmodeldir

Then you will get:

• saved model under resources/savedmodeldir/
• frozen graph with name resources/savedmodeldir/efficientdet-d0_frozen.pb

3. Export Feature Extractor frozen graph.

!rm  -rf resources/networks
!tar -zxvf resources/deep_association.tar.gz

4. Run The Tracker.

!python smart_ordering.py --tracker_model_name=deep_sort --image_size=512x512

Check python smart_ordering.py -h for an overview of available options.

5. Package Diagram.

├── resources
│   ├── deep_association.tar.gz
│   ├── deep sort.odt
│   ├── deep_sort.pdf
│   ├── efficientdet-d0.tar.gz
│   ├── efficientdet.pdf
│   ├── MOT_class_diagram.mdj
│   ├── Smart_Real_Time_Ordering_Paper.docx
│   ├── Smart_Real_Time_Ordering_Paper.pdf
│   └── sort.pdf
│
├── detectors
│   ├── detection.py
│   ├── detector.py
│   └── eff_det_0.py
│
├── trackers
│   ├── deep_tracker.py
│   ├── kalman_filter.py
│   ├── sort_tracker.py
│   ├── tracker.py
│   └── track.py
│
├── utils
│   ├── features_util.py
│   ├── hyper_params.py
│   ├── nn_matching.py
│   ├── overlay_util.py
│   └── util.py
│
├── smart_ordering.py
└── model_inspect.py

6. Highlevel overview of source files

In the top-level directory are executable scripts to execute, evaluate, and visualize the tracker. The main entry point is in smart_ordering.py. This file runs the program with front camera of laptop. In package detectors is the main detecting code:

• detector.py: Detector base class that represent a blueprint for other detectors models and should adopt it.
• detection.py: Detection base class.
• eff_det_0.py: child of detector.py and our offical tracker.

In package trackers is the main tracking code:

• tracker.py: Tracker base class that represent a blueprint for other trackers models and should adopt it.
• track.py: The track class contains single-target track data such as Kalman state, number of hits, misses, hit streak, associated feature vectors, etc.
• kalman_filter.py: A Kalman filter implementation and concrete parametrization for image space filtering.
• sort_tracker.py: implementation of SORT algorithm for tracking.
• deep_tracker.py: implementation of Deep_SORT algorithm for tracking.

In package utils is the main helper tool's code:

• features_util.py: This module contains code helps in build and run feature extractor model that used in deep_SORT algorithm.
• hyper_params.py: This module contains code for default parameter value.
• nn_matching.py: A module for a nearest neighbor matching metric.
• overlay_util.py: A module contains a low level drawing functions for drawing overlay bounding boxes over orignal image.
• util.py: This module contains helper code for min cost matching problem solving and the matching cascade algorithm and linear algebra operations.