reinforce Landslide warning system by private LoRa

According to the 2020 edition of the Global Climate Risk Index, a report that ranks the impact of weather disasters such as typhoons and floods, the worst is Japan and the second is the Philippines.

In the Philippines, many houses are handmade and buildings are not as strong as in Japan. Most families are not financially able to afford to build new houses after suffering damage from disasters. In addition, over the course of the 20th century, the forest area in the Philippines decreased from 70% to 20%.

 Illegal logging has occurred in the Philippines, and flood damage is becoming more severe in some areas.

Current infrastructure maintenance practices are limited to steady-state maintenance. Disaster prevention work has been inadequate and inappropriate, especially in sections with critical slopes, and improvement efforts have been limited to reopening infrastructure, resulting in repeated slope hazards and increased maintenance costs.

The method proposed in this study is not a matter of drastic solutions, but rather a system to quickly evacuate people in the event of an earthquake or a collapse, with the primary concern of saving lives!

 

LoRa system

cost

visualize

 

Difficulty in setting up M5STACKCORE for the combined use of LORA and GPS modules

I have considered using the LORA and GPS modules of M5STACKCORE together, but have not been able to get them to work well together yet.

With SPI communication, the pins other than SS are common, so the sensor used by the SS pin can be switched. However on M5Core2 by default GPIO33 is used as external I2C(SCL).

Core2 pin-array

https://raspberrypi.mongonta.com/m5stack-core2-news/

0) GPS unit

GPS unit

GPS unit

1) ES920LR3 by MaiaR Create

ES920LR3

①Tx-32,Rx-33/default : 33pin is in conflict, with LoRa module and GPSunit.

②Tx-13,Rx-14 : under consideration

③Tx-38,Rx-26 : under consideration

2) LoRa868 module

LoRa868 pin-array

33pin is in conflict, with LoRa868 module and GPSunit.

3) Module GPS SKU:M003

Module GPS SKU:M003 has not been purchased yet, but the possibility of using it with LoRA may be possible since the SCI port can be selected, but the cost is prohibitive, so I am considering purchasing it.

m5-docs (m5stack.com)

The series communicate protocol between M5Core and GPS is UART, physically connected via UART2 (GPIO16, GPIO17)

M5Stack Fire has occupied GPIO16 / 17 to connect with the PSRAM by default, it's conflict with TXD / RXD (GPIO16, GPIO17) of GPS module. Therefore, when using the GPS module with the M5Stack Fire, you might have to cut the TXD and RXD from GPS module and wire fly to another set of UART pin

SKU:M003

 

chatBingと論文解析

『BingのAIチャット機能』にもGPT-4が採用されていることが公表されました。無償利用ということで回数制限とAPIが公開されていないなどの制限があります。

 

【問い合わせ】EEGの解析について

 [注] 日本語の単語切り取り、あいまいさ、貯蔵データ量の問題から、法律などのローカライズの問題以外は英語で問い合わせる

chat Bing (GPT4.0model)

@hiroki_okuhata_int(NTTデータ先端技術)氏の「話題の ChatGPT + LangChain で、膨大な PDF ドキュメントの内容を爆速で把握」・「話題の ChatGPT + LangChain で ChatGPT が学習していない最新の OSS ソースコードを爆速でウォークスルーする」をもとに、技術文献の抽出抄録を入手する方法を備忘録として作成する。🩺

 

1)BingのAIチャット機能で、論文の抽出する。

** 調査サンプル **

Frontiers | Group-Level Multivariate Analysis in EasyEEG Toolbox: Examining the Temporal Dynamics Using Topographic Responses (frontiersin.org)

2)抽出文章のPDF ドキュメントからテキストを抽出する。

 オンライン/オフラインのファイルコンバーターで変換する

3)テキスト(元PDF) ドキュメントは、Google ドライブに格納して Google Colab から Python コードでアクセスする。

google drive 活用

4)分割したテキストの情報をベクターストアに格納する
 テキスト間の関連性についてのChatGPT に与えるベクトルデータをベクターストアに格納しておく、『DeepLake』 という Deep Learning 用のデータレイクサービスを使用します

DeepLakeにおけるDataSet例

5)LangChain における LLM のセットアップと
ChatGPT による自然言語でのクエリ処理を実行する

-> **Question** です: このレポートの目的は何ですか?1000字以内で説明してください。

LLM

摘要抽出

**回答**:このレポートの目的は何ですか?

本報告書の目的は、地形パターンの情報を用いて認知仮説を検証するための4つの多変量解析手法を含む手順を紹介することである。本報告書では、教師あり機械学習の枠組みでパターン分類法を用いて、すべてのタイムポイントにおける条件間の地形の違いを求め、地形の時間的変化を明らかにすることを述べている。また、全人的な地形情報を考慮した地形分散分析(TANOVA)とパターン分類を用いて、グループレベルの統計処理を行い、反応パターンの動態を調べることを紹介しています。

 

ESP32-Seismographs using Solar Cells

Unpowered seismographs are being considered using solar cells. WIFI is used for data transfer. Currently confirmed automatic measurement for about 2 weeks. LoRA will be considered to extend the transmission distance.

BigBlue 28W-SunPower Panel

on website Ambient

 

Generative Deep Learning by David Foster

CHATGPT is getting a lot of attention. I read again a clear enlightening book about its creativity and ran it. a modified sample that can be run on Google-Colab was available on the net, so I ran this.

BOOK


Since I was interested in composing, I ran the Compose chapter. The sample is Bach, whose geometric harmony and pious music both interest and attract me. However, the demo sample still needs improvement, but I think there may be a way to use it as a motif and for someone to arrange it.

VisualCode with Score4

calcurate-time 120min

 

ESP32 earthquake detector with solar battery

Experiments were conducted on the possibility of ITO devices based solely on solar batteries. The data will be uploaded to a server for viewing. The server was ambient, which is free of charge.

Sensors are monitored every second and uploaded every 20 minutes. However, if the deviation is large, it will upload every second.

Solar-charged mobile battery specifications 80x80mm 5000mAh 

ESP32 with ADXL345

Arudino IDE srccode

Duration 48 hours