---
tags: COTAI INTERNSHIP
---
# Internship - Nguyen Bao An

Day 1 (7th July)
1. Data science package: numpy, pandas
2. ML frameworks: scikit-learn
3. Python programming: function and class

Report for Day 1:
1. I have finished learning about the fundamental of numpy and pandas. With numpy, I know how to create an array, how to take index and slicing as well as performing some operations such as reshaping an array, stacking two different arrays, doing math operations on matrix. With pandas, I know how to create a series, dataframe, how to extract data from a csv file, and doing operations on dataframe such as concatenating different dataframes, cleaning N/A data, taking specific data and drawing graphs.
2. In Python, I have learned how to create and call a function (and anonymous function) as well as distinguish a local variable and a global variable and four different kinds of arguments: required argument, keyword argument, default argument, and variable-length argument.

Day 2 (8th July)
1. ML frameworks: sickit-learn: linear regression, classification
2. Python programming: function and class (cont.)

Day 3 (9th July)
1. Mathmatics review: linear algebra, probability and statistics

Day 4 (10th July)
1. Python progamming: class
2. ML frameworks: scikit-learn: classification
3. DS package: matplotlib

Report for Day 2, 3, and 4:
1. I have finished working with some supervised learning models in scikit-learn such as regressions (simple linear regression, multiple linear regression) and classifications (Naive Bayes, KNN)
2. In Python programming, I have finised studying about functions with some exercises. Also, I can plot some specific graphs using matplotlib package.
3. I have reviewd some basic mathematical fundamental of linear algebra, probability and statistics.

17th July:
1. Pytorch for deep learning (youtube) (1)

18th July:
1. CNN theory and model SSD (1)
2. Object detection (youtube) (1)

19th July:
1. Object detection (youtube) (2)
2. Pytorch for deep learning (2)

20th July:
1. Object detection (3)
2. Pytorch for deep learning (3)

21th July:
1. CNN theory: Model SSD (2)
2. Object detection (4)

22th July:
1. Object detection (5)
2. Segmentation theory
3. Some review on object detection: https://hackernoon.com/object-detection-theory-69a01db5aab4

23th July:
1. YOLO, ResNet theory (2)
2. YOLO theory: https://blog.paperspace.com/how-to-implement-a-yolo-object-detector-in-pytorch/

24th July:
1. Segmentation theory and pratice (1)
2. PSP theory: https://github.com/hszhao/PSPNet

25th July:
1. YOLO implementation on Pytorch (1)
2. YOLO basic introduction: https://www.youtube.com/watch?v=TxPimS50PU8
3. Segmentation (3)

27th July:
1. YOLO implementation (2)
2. Review on classification

28th July:
1. Review on SSD
2. Segmentation (4)

29th July:
1. YOLO implementation (cont.)

30th July:
1. Review on segmentation

3th August:
1. Modify the SSD and YOLO model
2. Applying real images

4th August:
1. Introduction to CNN with Keras: https://www.kaggle.com/yassineghouzam/introduction-to-cnn-keras-0-997-top-6

From 4th to 19th August:
1. Some modification with SSD, PSPNet, and YOLO and try other datasets (ADE20K, cityscapes).
2. More on SSD: https://medium.com/@jonathan_hui/ssd-object-detection-single-shot-multibox-detector-for-real-time-processing-9bd8deac0e06#:~:text=SSD%20is%20a%20single-shot,offsets%20to%20default%20boundary%20boxes
3. More on PSPNet:https://arxiv.org/pdf/1612.01105.pdf

From now, I will choose the PM5 and PM2.5 to predict the trending of dust in city area. For the data, I will measure the dust density in some community areas like district 1, 3, 10, and 11 as well as in suburb areas. From those data, I will construct a map of dust in HCMC and compare it to images from MODIS. After that, I will put the two datasets in a deep learning model to predict the trending dust.