Python Performance Optimization using Numpy, JIT, Cython, TaiChi, C++, OpenMP

This project's goal is to highlight the importance of using the right tools when the code is computationally intensive.
The code is designed to address a computer vision task that finds the True Positives (TP), False Positives (FP), and False Negatives (FN) in the COCO dataset.

Here, we first implemented a basic code on v1_python that has Evaluator class which reads two dataset in COCO format as ground truth json and prediction json and then finds all True Positives (TP), False Positives (FP), False Negatives (FN) and returns their corresponding annotation ids (in evaluate method).

Then we reimplement the same concept but utilizing tools that can give us performance boost:

• Rewriting code with numpy operations.
• JIT optimizations
• Optimizing code with tai_chi framework
• Cython classes, methods,no wraparound, no boundscheck
• Cython nogil and parallel processing, arrays and memoryviews
• Python integration with C++ and pybind
• C++ threading
• C++ Threading with Shared Mutex
• C++ OpenMP Optimization.

We tested the algorithms on 2 different sizes of jsons.

• Small size jsons: Almost 2K annotations.
• Big size jsons: over 99K annotations.

JSONS with 2K annotations:

version	name	process time (seconds)	RANK #
1	basic python	0.9084	9
2	numpy	2.5103	10
3	JIT	0.4453	8
4	Tai-Chi	0.1000	6
5	Cython	0.05395	5
6	Cython Parallel	0.13021	7
7	C++	0.00451	🥉
8	C++ Parallel	0.00175	🥇
9	C++ Parallel + Shared Mutex	0.00389	🥈
10	C++ OpenMP Parallel	0.01893	4

JSONS with 99K annotations:

version	name	process time (seconds)	RANK #
1	basic python	125	9
2	numpy	469	10
3	JIT	3.8314	6
4	Tai-Chi	2.2207	5
5	Cython	7.62609	7
6	Cython Parallel	18.65183	8
7	C++	0.63481	4
8	C++ Parallel	0.05023	🥇
9	C++ Parallel + Shared Mutex	0.53730	🥉
10	C++ OpenMP Parallel	0.05228	🥈