ESP32-CAM: Object classification with Tensorflow.js

[@atiradordeteia-brasil_01]

Please help. All tests I get only the prediction result: broom. HOW do I upload object images?

[@geofrancis2001]

are any identification strings sent back to the esp32? or is it just displayed on the android screen?

[@taylor.galbraith]

I am a little confused on the set of of the code. Do you have a tutorial going over setting it up?

[@maxencelancosme7594]

do you know why i can't connect to the wifi ? I use the esp32 like an acces point

[@nevzatyagmur1597]

lütfen projenin çalışması için açıklayıcı bir bilgi paylaşırmısınız adım adım işlemi anlatırmısınız

[@ryanjin8343]

Love this project man! Do you have any project (or know any other people's projects/tutorials) that runs a pre-trained Tensorflow model locally on the ESP32 CAM?

[@roboticslearner]

I think Tensorflow lite can be run on ESP32, you can check this page: towardsdatascience.com/tensorflow-meet-the-esp32-3ac36d7f32c7, or this www.tensorflow.org/lite/microcontrollers

[@atiradordeteia-brasil_01]

@@roboticslearner All tests I get only the prediction result: broom. HOW do I upload object images? +Subs

[@ConsultingjoeOnline]

_Awesome work!_ *SUBSCRIBED* Thanks!

[@sltechgalaxy1677]

please do a vedio for make this

Hi, it said connected to my wifi already. But, I can not access the IP address in Serial monitor. Could you help me??

[@jacobcook1450]

Hi! I'm really interested to utilise this code however I am a bit lost at how to implement, the readme on your github does not tell me so much about how to get this set up - I am using a Macbook Air if there are are differences to set up that I should be aware of? Please could you respond and let me know? Thanks so much - great work!

[@ashrafshorbaji2106]

My friend, how can I view it on my phone, where do I get the IP address for it?

[@MrTalhakamran2006]

Awesome, I want to build a plant vs weed identifier for garden :)

[@Artistlife4ever]

😂😂😂🦾

[@vgbhg9540]

How far did you go. This is my intention too.

[@MrTalhakamran2006]

@@vgbhg9540 unable to spare time from my work routine

[@TheJay6621]

Well done.. keep it up.

[@alejandromonsalvek.2014]

Awesome project, helped me a lot, thanks! A question regarding the prediction function. I'm trying to "automate" the onCommand and Predict functions, so instead of using the buttons, "onCommand" runs 1 time and then "Predict" once aswell. For some reason, pic is taken, but then, Predict function stops at "fromPixels()", and only once the Predict button is clicked, it'll run all the way through the code until the end. Any guess on why could this be? Appreciate your help!

[@ashrafshorbaji2106]

My friend, how can I view it on my phone, where do I get the IP address for it?

[@alejandromonsalvek.2014]

@@ashrafshorbaji2106 Hi Ashraf, you get the IP adress from the serial monitor through your Arduino IDE

[@tomming7729]

Awesome. Is it possiable to detect on real time?

[@roboticslearner]

I don't think it could be, we should use another approach and I am working on it, it would be uploaded soon

[@satriajulio7898]

heloo i want to ask I upload it with esp32 cam, it keeps failing, why? Does it use communication between esp32 and esp32 cam. love you know how

[@satriajulio7898]

how to upload program , I failed connection A fatal error occurred: Failed to connect to ESP32: Timed out waiting for packet headers

[@roboticslearner]

Hi satria, there are some rules to upload the esp cam module such as connect the gpio0 to gnd when reprogramming, you should check it by searching some tutorial of "how to upload code to esp32-cam"

[@satriajulio7898]

@@roboticslearner I have followed it, the problem is I uploaded your existing program but the Github link can't be uploaded failed to connect esp32

[@satriajulio7898]

may I ask for the original program that you created in the video

[@roboticslearner]

@@satriajulio7898 Sorry but what you mean original program ? And "can't be uploaded failed to connect esp32" means ? you mean your Arduino IDE cannot connect to esp32 ? If yes, it is because of your connection of esp32-cam module to your pc via an uart-usb module or your ide setting, I think.

[@rmsyauqi230]

hello i want to ask ,when I uploaded it to arduino it failed, that's why yah.failed to connect esp32

[@rmsyauqi230]

how to upload ?

[@roboticslearner]

Hi satria, there are some rules to upload the esp cam module such as connect the gpio0 to gnd when reprogramming, you should check it by searching some tutorial of "how to upload code to esp32-cam" on internet

Really impressive project! I have a couple of questions. - How does the classification itself happens? Is it done by the ESP? I dont find the relevant part in the .ino file. - I gave it a try and uploaded it to the ESP32-CAM. When I connect to the (correct) IP address I only see an empty website. (/capture works) Do you have any suggestions? Thank you in advance! Keep it up the awesome projects! Cheers:)

[@roboticslearner]

Hi David, I would answer your questions as following: 1. The classification happens in your web browser by running the the tensorflow.js (it is javascript code) , the ESP actually acts as streaming web server only, ESP actually support running tensorflow lite itself also, but maybe I will make another project for that later. 2. You should first flash the html (data) into the SPIFFS first, then flash the ino file as normal, then connect it again. I have updated README for it: github.com/ANM-P4F/ESP32CamClassificationTfjs/blob/master/README.md Please let me know if you still not run it.

Ahh, I missed that part! Thank you for the update, now it is working like a charm! However, the prediction is not too reliable, but it is still interesting to see what a low power device with the help of a web browser is capable of. Yeah, it would be interesting how the classification would perform on the ESP itself. I subscribed, so I will see if you end up with a video on this topic.

[@roboticslearner]

Hi Elena la Monaca, I think ESP can do some simple computer vision as color recognition, but for classification I don't think it has enough resource to do that, so I think ESP can be used mostly for streaming this moment ...

[@JamesT65]

@@roboticslearner thank you for the sample it works great on ES:P-EYE

[@JamesT65]

if anyone wants to know how to upload the data folder you can do the following Download and install the latest Python. When prompted by the installer, choose to add it to path. Log out and then back in for the changes to path to take effect. If you're using Linux, use the default package manager to install Python. For ESP8266 use ESP8266FS Plugin. Download the archive. There will be a folder in this archive called ESP8266FS. Put this folder in your sketchbook location, tools subfolder. Create this subfolder if it does not exist. For ESP32 use ESP32FS Plugin. Download the archive. There will be a folder in this archive called ESP32FS. Put this folder in your sketchbook location, tools subfolder. Create this subfolder if it does not exist. The sketchbook folder is by default in User's Documents, Arduino for Windows users. On Linux it is ~/Arduino. Launch Arduino IDE and if you placed the plugins in the correct directory you should have some new menu items in Tools.

[@alfredomaussa]

awesome

[@tanmaysharma7409]

can we use this for waste sorting??

[@roboticslearner]

Yes, but I think you should retrain your model with your own dataset for more accuracy, this pretrained model is for general using with many classes,..

[@manzoorahmed6048]

That's truly amazing man and I've some questions regarding this. 1. I'm working on my project that detects a real-time fire, does ESP32 works well in classifying the fire or should I consider Raspberry Pi? 2. How can we run the camera on an Internet server? Your videos are amazing, subscribed 💯

[@roboticslearner]

Esp32 is too slow for your project, you should use other kind of board, raspberry pi is OK I think

[@manzoorahmed6048]

@@roboticslearner Thank you for your quick response, would you please tell me how to run the Camera screen on the webserver

[@444haluk]

0.99 PERCENT? MANNN, those are very low probabilities, that's like 1/100 :D