Machine Learning and Computer Vision Engineer
Resume:
Sadegh Soleymanipour
No**, Taravat No* Alley, Taravat St, Farzan Quarter, Ghods, Tehran, Iran
+989********* - *******@*****.*** - https://github.com/sasolp/ https://www.linkedin.com/in/sadegh-soleimani-29758466/ Summary
For 10+ years, I’ve been programming professionally in the field of Computer Vision Engineering. I have good experience in designing, training, and evaluating varied deep neural networks such as a deep network for License Plate Detection and Recognition, Low-Quality OCR, Lung Cancer Detection, Logo Recognition, Face Recognition, etc. It is worth noting that we achieved high accuracy in all of them. I have implemented and develop deep networks in many libraries such as Caffe, Tensorflow, Torch, Keras and MatConvNet. Before all of them, I had a great deal of experience in traditional methods used in the field of Image processing.
Skills
o Performance tuning and debugging experience (perhaps greatest Strength) o Self-driven problem solver
o 2D and 3D computer vision
o Object detection and Tracking
o Linear algebra, mathematics and statistics
o Strong Machine Learning skills (SVM, NN, DT, Clustering methods) o Excellent Knowledge and Experience of Deep Learning o Strong C/C++/CUDA/Python/C#/Matlab programming skills o Good experience with Yolo models, Mask-RCNN, Faster-RCNN o Tuning AlexNet, GoogleNet, ResNet, Dense Net, Inception Nets etc. o Strong understanding of Data pipelines, API workflows, CI/CD o Excellent Knowledge and Experience of:
o OpenCV
o Tensorflow
o Caffe
o Torch
o Keras
o Scikit-learn(sklearn)
o Scikit-image(skimage)
o SQL
o DeepStream
o RaspberryPI
o Jetson Series
o Computer Vision Applications:
o OCR
o Face Recognition
o License Plate Recognition
o Logo Recognition
o Official Document Categorization
o Lung Cancer Detection
o OS
o Linux
o Windows
o Teamwork platforms
o Github
o Gitlab
o Jira
Experiences
07/2019 to Current Senior Computer Vision and Machine Learning Engineer Kaman Inc. - Tehran, Iran
o Key Responsibilities:
Led and directed a developer team involved 5 people.
A high accurate License Plate Detection And Recognition Model
Tracking, Classification, and Counting of Vehicles
Synthesizer for the Iranian license plates
The plates from various aspects and sizes
o Key Achievements:
Achieved an accuracy about of 98% for LPR
A C++ Library with 128 fps on GPU GeForce 1080 for LPR
Achieved an overall accuracy above 95% for counting of Vehicles.
Achieved an increasing in accuracy (+5%) by using the plate synthesizer
04/2014 to 06/2019 Senior Image Processing Engineer and Researcher Research Center For Development of Advanced Technologies-Tehran, Iran o Key Responsibilities:
Led and directed OCR, Face Recognition and Video Enhancement Projects with more than 3 people.
An OCR Model Based on composed method based on HMM and SVM
Two algorithms to Detect the Pornographic images
A Full Platform to Detect the Forgery images
A Dense Features extraction with CUDA (improved to 400 times faster)
Designed a C++ Engine for Face Detection and Recognition
Originated a new method to OCR the Low-Quality plate images
Utilized a complex composition of deep networks (VDSR and LeNet)
Consulted a few Projects related to deep learning (LSTM)
Contributed to a speech project based on a RNN model named DeepSpeech using Tensorflow in C++
o Key Achievements:
Achieved an accuracy about of 99.5% for OCR.
Achieved an accuracy about of 97% for Pornographic Detection.
Achieved an accuracy about of 98.75% for a dataset contained about 4000 individuals for Face Recognition.
Achieved an accuracy about of 81.5% for very low quality images of the license plates.
Achieved an accuracy about of 87.7 for OCR of the Low-Quality plate images.
Achieved an efficient implementation of DeepSpeech about 200 times faster than CPU by GPU GeForce 1060
06/2011 to 03/2014 Image Processing Engineer
Behin Pajouhesh Eng. Co - Tehran, Iran
o Key Responsibilities:
Designed and developed a high accurate method to segment various areas of the fingerprints froms.
An algorithm to determine the quality of fingerprint
An algorithm to determine the center of a fingerprint
An algorithm to extract the effective fingerprint area from non-quality images extracted from WSQ images
A new successful algorithm for In-painting of missed area of the direction image.
o Key Achievements:
Achieved an accuracy of over 98% in total above-mentioned Projects.
04/2008 to 08/2010 Image processing engineer
Partogostar Novin - Tehran, Iran
o Key Responsibilities:
Designed and Developed an OMR (Optical Mark Reader)
Reading any kind of OMR forms
Developed a GUI for designing, printing and automatic reading of OMR using Visual C#
o Key Achievements:
Achieved an accuracy of over 99%.
Additional Information
o I am ready to relocate immediately after receiving the job offer and then completing the task related to work permit.
Education
o Master of Science: Software Engineering, Payame Noor University - Tehran, Iran o Bachelor of Science: Software, University of Applied Science and Technology - Fouman, Iran
Publications
"A New Approach For Detection And Localization Of Duplicated Regions In A Forgery Image Based On Matching The Statistical Features Of Ring Blocks", 7th on Image Processing and Sight Machine.
S. Soleimanipour, A. Jodeiri, H. Rashidi, M. Mirhassani, H. Kheradfallah, H. Seyedarabi, “Automatic Ship Classification Utilizing Bag of Deep Features,” arXiv:2102.11520 [cs.CV], Feb. 2021.
These Will submit as soon as possible:
o “Lung Cancer Detection using Encoded Deep Features” o “Low Quality License Plate Recognition by Composed Inceptions of VDSR”
Contact this candidate