Intern AI/Machine Learning Engineer

EDUCATION

*rd student of Computer Science at UIT – University Information

Technolory

GPA : 8.12 / 10.0

Certification:

• Convolution Neural Networks (Coursera)

• Neural Networks and Deep Learning ( Coursera)

• Structuring Machine Learning Projects (Coursera)

• Improving Deep Neural Networks: Hyperparameter

Tuning, Regularization and Optimization (Coursera)

• Sequence Models (Cousera)

CertificationLink:https://drive.google.com/drive/folders/1uMWSsT6HoUi S2r_QLJJsf1UhnCaZN5vS?usp=sharing

Course:

• DeepLearning ( Coursera) – Full Course

Programming Languge: C++, Python

Libraries: Scikit-learn, TensorFlow,

NumPy, Pandas, Matplotlib, Keras

Others: Github, Google Colab, Visual

studio

Lê Phước Đạt

AI Engineer

TECHNICAL SKILLS

Participating in school algorithm contest : WECODE_CHALLENGE2018 Participating in school algorithm activity: ALGO BOOTCAMP2018 Participating in school AI contest: AI Tempo Run 2021 Encouring learning scholarship at 2nd semester of the academic 2018-2019 OBJECTIVES

Name : Lê Phước Đạt

Date of Birth : 8th Sep 2000

Address :

National University A dormitory, 6 Town,

Linh Trung Ward, Thu Duc City, Ho Chi

Minh CT

Phone : +849********

Email: **************@*****.***

• Get into a large company, work under a

professional enviroment

• Contribute your ability to projects

• Become a leading expert in AI industry

• Bring value to society

• Knowledge, good programming ability

• Have programming thinking,

good ability to solve problem

• Actively learn, research and seek

new technologies

• Adapt quickly to new enviroments

• Good interpersonal and

communication skills PROJECTS

SKILLS

ACTIVITIES AND AWARDS

PERSONAL INFORMATION

1, Gender Classification

- Description : Input: image, output: Gender

- Method: + Data : 2000 images from CelebA dataset

+ Use MTCNN to face alignment

+ Reduce dimension with pretrained facenet

+ Train in multi Machine Learning model, tuning

hyperparameter and select the best

+ Enhance 200 images about elderly and children and retrain to overcome the disadvantages of CelebA dataset

- Team size: 1 members

- Role : Do everything

- Finishing time : 2 months

- Tech used : Python, Sklearn, Keras

- Github: https://github.com/lephuocdat2000/-CS114.K21-

/tree/master/GenderClassification

2, Removing background with Chroma Key Technique

- Description: Input: image or video and background Output: Image or video which has been replaced background 5, Speech Emotion Recognition for Vietnamese

- Description : Input : Audio or Record for 5-10

seconds

Output: Emotion

- Method : + Collect 500 Vietnamese audio and

combine with IEMOCAP and RAVDESS dataset

+ Build model using two methods:

traditional Machine Learning with hand-crafted

features ( pitch, harmonic,…) and Deep Learning

+ Conduct three experiments for

each method:

- Use only Vietnamese dataset

- Use English dataset to train

then test on Vietnamses dataset

- Combine two languages

+ Evaluate two methods and choose

the best one

- Team size: 5

- Finishing time: 1 months

- Role: Build Deep Learning model and conduct

three experiments

- Tech used : Python, Flask

- Github:

https://github.com/lephuocdat2000/Speech-

Emotion-Recognition-for-Vietnameses

PROJECTS

- Method: + Convert image or frame (video) from RGB to HSV color space

+ Detach H and S color space and use K-means Clustering algorithm to cluster background pixel value in these color spaces

+ Calculate Gauss distribution of background

+ Create image mask with Gauss distribution

+ Remove background with formula: mask * image +

background * (1- mask)

- Team size: 3 members

- Role : Propose solution, Back-end

- Finishing time: 1 months

- Tech used: Python, Sklearn, Flask

- Github:

https://github.com/lephuocdat2000/NhapmonCV/tree/master/%C4%90%E 1%BB%93%20%C3%A1n%20Th%E1%BB%8B%20gi%C3%A1c%20M%

C3%A1y%20t%C3%ADnh

3, Part - of Speech tagging for Vietnamese ( POS)

- Description : Input : Vietnamese sentece

Ouput: Sentence has been seperated from word and

label with each word

- Method: + Collect 60 Vietnamese sentences

+ Word separation with Vietokenizer

+ Labeling data using website:

https://vlsp.hpda.vn/demo/?page=vcl

+ Build model with Hidden Markov algorithm

- Team size: 3 members

- Role: Participating in collecting data, build model

- Finishing time : 1 months

- Tech : Python, Flask

- Git:

https://github.com/lephuocdat2000/NLP/tree/main/Pos%20Tagger%20for

%20Vietnamese

4, Social Distancing Violation Detection

- Description : Input: image or video

Output: image or video with red bouding boxes if it has violation

- Method: + Pick 4 points to define detection area ( ROI )

+ Use pretrained Yolov4 to detect person and choose bottom

– center point in each bouding box to represent that box coordinate

+ Convert ROI to top-down view with perspective transform

+ Calculate scale beetween pixel and real distance

+ Calculate pixel distance beetween each pair of bouding boxes and using scale to convert it to real distance

+ Drawn red boxes for pairs that have distance under 2m

- Time size: 2 members

- Finishing time: 2 months

- Role: Participating in contributing solution, coding, optimzation and model validation

- Tech used: Python, Flask

- Github: https://github.com/lephuocdat2000/Advanced- CV/tree/main/Social%20Distancing%20Violation%20Detection

Contact this candidate