2019 Mobile Thermal App Challenge

Blog

We are applying for 2019 Mobile Thermal App Challenge for FLIR products.

We started to play with FLIR Lepton thermal data WiFI transmitting ESP8266 to iPhone in 2017.

サーモドローン flir lepton 1 + wroom02(esp8266) + furibee F36(jjrcH36) + iPhone apps
WROOM02+LEPTON1 (80x60画像)をドローンに取り付けてiPhoneで受信。サーモ画像は作成した専用アプリの録画機能を使用。これでmaker faire 2017 に参加します。Take's diaryのTAKEsanと一緒にW&Tとしてjetson TX2を駆使してこの画像からディープラーニング...

We spent time and money for try and error for software and hardware a lot since we are not professional for this kind.

for my business, we check our own PV sight with this system.

太陽光 メンテナンス ホットスポット サーモドローン(dobby 199g + flir lepton3 attachment 28g= total 227g)
無人航空機の重量については、機体本体の重量及び飛行のために必要なバッテリーの重量の合計(バッテリー以外の取外し可能な付属品の重量は含まない。hubsan h502はwroom02と混線してノーコンになります。赤外線センターアタッチメントは軽量ですが混線して誤作動するモデルがあります。

it was fine with us but my friend, who is playing with deep learning combined with Yolo with this data(coco data). it works like this.

deap leaning(AI)
YouTube でお気に入りの動画や音楽を楽しみ、オリ...

using 2sets of Lepton system, we also made prototype stereo thermal cam. and applied for FLIR Lepton Maker challenge runner-up and listed lepton developer showcase. Thanks Cal-san !

Why we are developing this ? to start with wireless, handy and ultra-light system for thermography. now adding deep learning with it seriously.

now we are making Pre-learned models for thermography data.

Although we try with FLIR Pre-learned model but not really work well, we add original Pre-learned data and we are making our original system, wireless, handy and ultra-light system with Yolo v3 not only Lepton but also BOSON, which based on our Lepton system.

Upto September 2019, we have been using NVIDIA Jeston(TX2, Xavier and Nano), which never be handy but we have completed our system Toy ADAS.

Jetson Nano+Lepton3.0+tiny YOLOv3
The real-time image of LEPTON 3.0 sent from ESP 8266 was used to identify cars, people, pedestrian crossings and bicycles using Jetson nano. I use tiny YOLOv3.

protetype with 3D printer.

I have chance to show this system to FLIR marketing manager Mr.Greg Nagler in Tokyo this may.

He advised me this system to promote BtoC, such as developers and high-amateur but not BtoB.

and name should be Kansai Thermal Cam never be Thermal camera, which will be hit full of Chinese products and can not find our system, who is looking for our system via web.

this challenge is also our greatest opportunity to show our systems many developers, who want to use thermography with deep learning.

For business ? money ?, We are playing with Lepton and BOSON with Fun and share with others(not too many pull request).

Jetson Xavier + FLIR Boson QVGA 60Hz
FLIR Boson QVGA 60fps 画像で YOLOv2のジャンケン認識

at the moment, we found it difficult to combine with FLIR developer kit, which is condition of this challenge.

Rather Apply rise more difficulties to apply for iOS apps with this system at the moment.

never mind we will play with them more anyway.

Now a professor ask us to develop to detect elderly fall down with thermal cam system at hospitals or care house.

so we are working with him at the moment.

Why thermal cam ?

Cornell universal are studying this kind with a stereo cam but his point is not to use camera due to privacy issue but thermal cam can not identify the person but can detect elderly fall down there.

 

We are trying to clear apple regulations at the moment and need some more time.

Stay tuned !

コメント

タイトルとURLをコピーしました