Blog¶

Jun 13, 2021
in DevOps
1 min read

Draft

Deep Learning

AI PPTs

This article collects excellent PPTs on AI.

The future of AI- Lecun

May 13, 2021
in AI, Language Model
1 min read

Draft

NLP- Word Understanding

Overview

[![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4440537%2Fa3a11ff3167936d62cfc8af32

May 11, 2021
in Data
1 min read

Draft

Tracking vs. Privacy

References

《選擇同意(opt-in)？選擇退出(opt-out)？企業因應兩種不同制度設計注意事項》（台灣）

Apr 28, 2021
in 3D
1 min read

Draft

AIOT data collection- Part 2

This post focuses on concepts of big data before go deeper of technology.

Bigger picture of Big Data

Better service cycling from a deeper insights of data

Deeper insights of data

Deeper insights of data: Coming from evolution of data analytics service

Deeper insights of data: Coming from more automation

Apr 10, 2021
in Data, IOT
2 min read

Draft

AIOT is the combination of AI and IOT. Three major steps for AIOT: data, intelligence, actions mean data collection, smart data intelligence, then actions accordingly. For data collection, IOT devices events and user behaviorial events might be collected via several approaches. Here we focus on the approaches of data collections and aggreagration.

via InfluxDB

Data collection, or IOT devices and user behaviorial events collection, can be formulated as time series or events and and stored in times series database (TSDB)

a time series or event can be in the form of

1	`measurement tag fields timestamp`

andd an example as

1	`temperature,device=device1,buidling=b1 internal=80,external=18 1443782126`

influxDB is the top solution for (TSDB) as collectors, aggregators and visualizer and its block digram as below.

A showcase of IOT data collection architecture via influxDB.

A detailed pipeline for telegraf plugins: Inputs, outputs, processors plugins to process the data.

An example show how FLUX to handle the data source buckets, pipe, and forward operator, and tables.

from(bucket:"mongo")
  |> range(start:-1h)
  |> filter(fn:(r) =>
    r._measurement == "cpu" and
    r.cpu == "cpu-total"
  )
  |> aggregateWindow(every: 1m, fn: mean)

References

Feb 23, 2021
in AI, Language Model
5 min read

Draft

Customer Feedback Tagging with NLP

Overview

Project

With the data pipeline below to collect, pre-process, feature-engineer, NLP alogorithm applied to provide useful dashboard for analysis and further actions.

Steps

Data Collection:

Data mining or ETL (extract-transform-load) process to collect a corpus of unstructured data.

Data Preprocessing:

Tokenization: Segmentation of running text into words.
Lemmatization: Removal of inflectional endings to return the base form.
Parts-of-speech tagging: Identification of words as nouns, verbs, adjectives etc.
Lanaguage detection: Identification of the lanauges from single or several sentensces even a short one.

Feature Engineering (NLP visualization):

Word Embeddings: Transforming text into a meaningful vector or array of numbers.
N-grams: An unigram is a set of individual words within a document; bi-gram is a set of 2 adjacent words within a document.
TF-IDF values: Term-Frequency-Inverse-Document-Frequency is a numerical statistic representing how important a word is to a document within a collection of documents.

Application of NLP Algorithms:

Latent Dirichlet Allocation: Topic modeling algorithm for detecting abstract themes from a collection of documents.
Support Vector Machine: Classification algorithm for detection of underlying consumer sentiment.
Long Short-Term Memory Network: Type of recurrent neural networks for machine translation used in Google Translate.

Scope

Topic Modeling: How to automatically categorize customer complaints or intent classification?

Sentiment analysis
- How to detect sentiment from customer feedback, a complaint or a positive feedback?
- How to detect urgency?

Implementation: WinkNLP

Customize tagging keyword

Implementation: spaCy

Customize tagging keyword

A high level view of generic model and the refine model in the whole process.

The detailed NLP refinement model is as below to improve the models of NER Tagging in spaCy model on user feedback.

Another better idea is Active learning as below

and the whole data pipeline diagram for user feedback tagging is as below

References

spaCy NER

NLP Kits

Label Annotation

ML Backend

Sep 18, 2020
in AI, Computer Vision
1 min read

Tensorflow.js POC 13: Avatar Generator with Face-API.js

Overview

Avatar Generator with Face-API.js

This POC, a consequential POC of face-api.js, regonize the face from camera and find the nearest avatar from thousands avatars generated from avatar generators.

References

Sep 15, 2020
in Data, IOT
10 min read

Pytorch POC 6: WIFI Indoor Positioning

Project Map of DLC

Overview

This is still a very early stage of POC. With a pre-defined datasets, with many WIFI APP signal strengths data try to prdict the real location of the indoor room.

This might be the first POC try to define our input and output data formats.

Also, this POC also be a high level connector of several different components together since it will leaverage multi factor information to conclude a precise indoor locationing.

When it comes to localization within buildings, a distinction can be made between client-based (Active tracking) and server-based positioning ( Passive tracking + Active tracking as optional). Client-based localization enables determining the position directly on the end user's device (e. g. smartphone). In the case of server-based localization, positioning takes place on a server.

Technologies for Client-Based Indoor Positioning

Client-Based Indoor Positioning Compared: Wi-Fi vs. BLE vs. UWB vs. RFID vs. Ultrasound

Technology	Accuracy	Range
WIFI	<15m	<150m
BLE 4	<8m	<75m
BLE 5.1	<1m	<75m
UWB	<30cm	<150m
RFID	Present detection only	<1m

Technologies for Server-Based Indoor Positioning

Server-Based Indoor Positioning Compared: Wi-Fi vs. BLE vs. UWB vs. RFID vs. Ultrasound

Technology	Accuracy	Range
WIFI	<15m	<150m
BLE 4	<8m	<75m
BLE 5.1	<1m	<75m
UWB	<30cm	<150m
RFID	Present detection only	<1m
Ultrasound	<4m	<8m or Wall

Indoor Positioning Technology	Comments
Client-Based	No server setup. Use existing framework. but, low precision
Server-Based (Passive tracking only)	Server only. 3-5 sec delay. Not realtime application
Server-Based (Passive+Active tracking)	Need sever+client to get hybrid info for realtime & stable location info

Git Repo	Status	Progress	Comments
			Family=barco API:Devices by Loc (w 1 mins)

Get info of device='dlc2.barco.com' & family='barco' from find3 server

curl -s -L -XGET "http://dlc1.barco.com:8005/api/v1/location_basic/barco/dlc2.barco.com"

Get WIFI/BT SSID, BSSID, RSSID of the location and device

1	`curl -XGET http://dlc1.barco.com:8005/api/v1/location/barco/dlc2.barco.com`

Get all devices within 1 minsfrom family='barco' from find3 server and output it as a table

curl -s -L -XGET "http://dlc1.barco.com:8005/api/v1/by_location/barco?history=1&num_scanners=1" | jq -s '.[] | .locations | .[] |.devices' | jsoncsv -A  | mkexcel | csvcut  -d , -C 5  |csvlook --snifflimit 0  | sudo tee output.txt

Update GPS info of location 'Aris Cube' (device=dlc2.barco.com)

curl -s -L -XPOST "http://dlc1.barco.com:8005/api/v1/gps" -H "Content-Type:application/json" --data-binary '{"f":"Barco","l":"Aris Cube","gps":{"lat":25.01290,"lon":121.46701,"alt":0}}'

Delete family=barco data (Dangersous operation!!! Warning!!!)

 curl -s -L -XDELETE "http://dlc1.barco.com:8005/api/v1/database/barco"

Block diagram of find3

Signal strength (dBm)	Expected Quality
-90	Chances of connecting are very low at this level
-80	Unreliable signal strength
-67	Reliable signal strength– the edge of what Cisco considers to be adequate to support Voice over WLAN
-55	Anything down to this level can be considered excellent signal strength.
-30	Maximum signal strength, you are probably standing right next to the access point.

A snapshot of find3 active RSSI & its BSSID, SSID on 2020/08/31 nearby dlc2.barco.com

Rssi	Freq	Type	Algo	Rate	BSSID	SSID
-50	2.412 GHz	802.11n	6	12	E8 D0 FC BF 18 0D	ClickShare-Mobile-CX20
-52	5.825 GHz	802.11ac	7	8	04 D4 C4 D3 42 84	ASUS_Automation_5G
-72	5.825 GHz	802.11ac	7	8	E4 AA EA 58 43 93	CX-50-DEMO-RICHEN
-50	2.457 GHz	802.11n	7	12	B0 7F B9 82 03 04	NETGEAR25
-66	2.437 GHz		7	7	F0 1D 2D 5D A9 C0	Barco
-58	5.24 GHz	802.11n	6	8	28 24 FF 69 5B B7	TestRoom_CSE800_mac (5GHz)
-55	2.437 GHz		7	7	F0 1D 2D 5B 54 61	Barco Guest
-61	5.24 GHz	802.11ac	1	8	3C 91 80 84 95 07	ClickShare-HW-UniSee
-71	5.24 GHz	802.11ac	7	8	E4 AA EA 35 04 65	ClickShare-1863551994
-55	2.437 GHz	802.11n	7	12	BE 42 4F CB C3 44	TestRoom-EAP-WinServer
-46	5.785 GHz		6	8	7C 10 C9 61 97 84	TAI-QA-ASUSROG-6E_5G
-49	5.18 GHz	802.11ac	7	8	10 63 C8 97 03 9F	TAI_MR08
-58	5.785 GHz	802.11ac	7	8	BC CF 4F CB C3 43	TestRoom-EAP
-63	5.18 GHz	802.11ac	6	8	3C 91 80 84 9C 5D	ClickShare-1862300001
-59	5.3 GHz		7	6	F0 1D 2D 5A D5 6F	Barco
-57	2.412 GHz	802.11n	7	12	D4 6E 0E 41 61 EA	keroro24
-60	5.32 GHz		9	6	F0 1D 2D 5B 54 6E	Barco Guest
-68	5.58 GHz		7	6	F0 1D 2D 5A D9 8D	BarcoIoT
-55	5.18 GHz	802.11ac	7	8	3C 91 80 84 9C 5F	TAI_MR09
-59	5.745 GHz	802.11ac	7	8	E4 AA EA 74 46 FF	ClickShare-Mobile-CX50
-53	2.412 GHz		7	7	F0 1D 2D 5A D5 62	BarcoIoT
-69	5.805 GHz	802.11ac	6	8	D4 6E 0E 41 61 E9	keroro5
-62	5.24 GHz	802.11ac	7	8	F8 A2 D6 6E 09 CF	TestRoom_CSE200P_mac
-58	5.785 GHz	802.11ac	7	8	BE 43 4F CB C3 45	TestRoom-EAP-WinServer
-73	5.68 GHz		7	6	F0 1D 2D 5C 27 CF
-39	2.447 GHz	802.11n	7	12	2C FD A1 CD 32 38	ASUS
-69	5.765 GHz	802.11ac	7	8	70 2E D9 42 F4 76	MAXHUB-6BF
-56	5.745 GHz	802.11ac	7	8	E4 AA EA 35 21 3F	SClickShare-1863551600
-58	5.785 GHz	802.11ac	7	8	BE 43 4F CB C3 44	TestRoom-PSK2
-62	2.412 GHz	802.11n	7	12	E8 D0 FC BF 13 5F	ClickShare-1862300251
-59	5.18 GHz	802.11ac	7	8	E4 AA EA 74 2C D3	ClickShare-1863553421
-60	2.452 GHz	802.11n	7	12	04 D4 C4 35 69 E8	ASUS_SQA_JessicaHsu_2.4G
-59	5.18 GHz	802.11ac	7	8	3C 91 80 84 95 D3	ClickShare-MR12
-69	2.412 GHz		1	7	F0 1D 2D 5A D9 82	BarcoIoT
-61	2.412 GHz	802.11n	7	12	BC 30 7E F1 1D 9B	ClickShare Audi-1
-61	5.18 GHz	802.11ac	7	8	3C 91 80 84 97 59	ClickShare-1862300078
-60	5.32 GHz		7	6	F0 1D 2D 5B 54 6F
-78	2.437 GHz	802.11n	7	8	02 21 6A F8 2F AF	DIRECT-ZWTAICLT24036msZe
-62	5.24 GHz	802.11ac	1	8	02 12 5F 17 67 2F	WiCS-2100-72E
-59	2.462 GHz	802.11n	7	12	00 D0 41 DC 13 C8	myvita
-55	5.18 GHz	802.11ac	7	8	3C 91 80 84 9F 25	ClickShare-0716174986
-71	5.18 GHz	802.11ac	7	8	10 63 C8 A7 9E 23	TAI_MR04
-86	5.18 GHz	802.11ac	1	8	10 63 C8 96 FF 87	TAI_MR01
-28	2.427 GHz	802.11n	1	12	7A DA 88 B2 74 EC
-66	2.422 GHz	802.11n	7	12	BC 30 7E DD 3D FE	WiPG-1000-78C
-56	5.18 GHz	802.11ac	7	8	F8 A2 D6 8D BF 8F	ClickShare-1863550102
-59	5.3 GHz		7	6	F0 1D 2D 5A D5 6D	BarcoIoT
-73	5.24 GHz	802.11n	7	8	D8 61 62 8B 1D EF	TAI-MR06-Pingtung
-52	5.18 GHz	802.11ac	7	8	12 63 C8 14 92 3B	DIRECT-BM
-83	5.68 GHz		7	6	F0 1D 2D 5A CC AE	Barco Guest
-81	5.26 GHz		7	6	F0 1D 2D 5D B2 AF
-68	2.437 GHz	802.11n	7	12	BC EE 7B 7D 46 C0
-81	5.745 GHz	802.11ac	7	8	02 12 5F 17 68 8C	WiCS-2100-88B
-68	5.58 GHz		7	6	F0 1D 2D 5A D9 8E	Barco Guest
-55	2.437 GHz	802.11n	1	12	BE 42 4F CB C3 43	TestRoom-PSK2
-77	5.745 GHz	802.11ac	7	8	02 12 5F 17 66 66	WiCS-2100-665
-52	5.18 GHz	802.11n	7	8	28 24 FF 4D 1B FD	ClickShare-1872075087
-82	5.2 GHz	802.11n	1	8	BC 30 7E D9 C3 F4	ClickShare-8000000049
-72	5.18 GHz	802.11ac	7	8	E4 AA EA 74 26 59	CX-50_Service
-71	5.18 GHz	802.11ac	7	8	E8 D0 FC BF 15 D1	Agile 01
-58	5.18 GHz	802.11ac	7	8	3C 91 80 84 95 B3	ClickShare-1862300131
-48	5.745 GHz	802.11ac	7	8	2C FD A1 CD 32 3C	ASUS_5G
-47	5.18 GHz	802.11ac	6	8	E4 AA EA 58 45 8B	ClickShare-1863552495
-84	5.22 GHz	802.11ac	7	8	3C 91 80 84 98 13	TAI_MR03
-69	5.18 GHz	802.11n	7	8	B8 B7 F1 01 B0 2D	ClickShare-1873124000
-45	2.412 GHz	802.11n	1	12	C8 60 00 AC FB 30	ASUS-FA
-64	2.462 GHz		1	7	F0 1D 2D 5C 27 C0
-68	5.58 GHz		7	6	F0 1D 2D 5A D9 8F	Barco
-54	2.412 GHz		7	7	F0 1D 2D 5A D5 61	Barco Guest
-64	5.745 GHz	802.11ac	7	8	02 12 5F 17 67 F2	WiCS-2100-LIN
-53	5.2 GHz	802.11ac	9	8	04 D4 C4 35 69 EC	ASUS_SQA_JessicaHsu_5G
-64	2.462 GHz		7	7	F0 1D 2D 5C 27 C2	BarcoIoT
-59	5.18 GHz	802.11ac	7	8	F8 A2 D6 8D BF 5D	ClickShare-1863550098
-81	5.745 GHz	802.11ac	7	8	02 12 5F 30 00 AC	WiCS-2100-0AB
-53	2.462 GHz	802.11n	7	12	00 4E 35 1A C8 60	TAI-ClickShare-WPA2-DFS
-73	5.68 GHz		9	6	F0 1D 2D 5C 27 CD	BarcoIoT
-59	5.18 GHz	802.11ac	7	8	F8 A2 D6 8D CB 39	ClickShare-1863550140
-75	5.22 GHz	802.11n	7	8	28 24 FF 5B 05 05	CSE-200-demo
-61	5.18 GHz	802.11ac	7	8	10 63 C8 BF B1 81	ClickShare-1862337967
-80	5.26 GHz		7	6	F0 1D 2D 5D B2 AE	Barco Guest
-53	5.18 GHz	802.11ac	1	8	A0 40 A0 82 49 0E	QA-test-5G
-51	2.412 GHz	802.11n	7	12	10 63 C8 96 F7 F3	ClickShare-T20
-60	5.32 GHz		7	6	F0 1D 2D 5B 54 6D	BarcoIoT
-43	5.22 GHz		7	8	6C CD D6 F5 9D 69	HW_WIFI6E_5G
-52	2.462 GHz	802.11n	7	12	18 31 BF C5 D1 38	SQA_Balloon_24G
-82	5.68 GHz		7	6	F0 1D 2D 5A CC AF	Barco
-84	5.745 GHz	802.11ac	7	8	02 12 5F 30 06 4F	WiCS-2100-64E
-45	2.437 GHz		7	12	7C 10 C9 61 97 80	TAI-QA-ASUSROG-6E_2.4G
-58	5.18 GHz	802.11ac	7	8	E4 AA EA 58 43 81	ClickShare-1863552454
-75	5.5 GHz		7	6	F0 1D 2D 5D A9 CD	BarcoIoT
-75	5.18 GHz	802.11ac	7	8	D8 F3 BC 54 4B 89	ClickShare-1862375851
-43	2.472 GHz		7	12	08 36 C9 2F 78 F5	TAI-QA-NetGear-2.4G
-63	2.412 GHz	802.11n	7	12	3C 91 80 84 9C 91	ClickShare-1862300102
-68	2.412 GHz		6	7	F0 1D 2D 5A D9 80	Barco
-66	2.437 GHz		7	7	F0 1D 2D 5D A9 C2	BarcoIoT
-63	5.5 GHz	802.11ac	7	8	00 4E 35 1A C8 70	TAI-ClickShare-WPA2-DFS