Video není dostupné.
Omlouváme se.
How to do Object Detection using ESP32-CAM and Edge Impulse YOLO Model
Vložit
- čas přidán 15. 08. 2024
- For Code and Circuit: circuitdigest....
In this video, we demonstrate how to build an image recognition system using the ESP32-CAM module to identify vegetables like tomatoes, potatoes, and onions. Using the Edge Impulse platform, we train and deploy a machine learning model to run directly on the ESP32-CAM. The results are displayed on an OLED screen for easy viewing. Learn the basics of edge computing, data acquisition, model training, and deployment. Check out our detailed step-by-step guide to create your own object detection system!
Hashtags: #ESP32CAM #EdgeImpulse #AIonEdge #ObjectRecognition #MachineLearning #IoT
Long Tail Keywords:
How to use ESP32-CAM for image recognition
Training AI models with Edge Impulse
Vegetable detection using ESP32-CAM
Deploying machine learning models on ESP32-CAM
Edge computing with ESP32-CAM and Edge Impulse
-------------------------------------------------------------------------------------------------------------------------
Please Like, Share, and Subscribe to our CZcams Channel: www.youtube.co...
To access 1000+ DIY projects and circuit designs, visit: circuitdigest....
You can visit circuitdigest.... to start a discussion with our engineers regarding any project
Subscribe to our Newsletter to never miss an update on CircuitDigest:
eepurl.com/dJ5xqU
Find the latest electronics industry news covering new electronics components releases in various electronics domains such as embedded, power electronics, analog, digital, IoT, and Automotive industries: circuitdigest....
-------------------------------------------------------------------------------------------------------------------------
Connect with CircuitDigest:
👉 Website: circuitdigest....
👉 Instagram: / circuit_digest (@Circuit_Digest)
👉 Telegram: t.me/joinchat/...
👉 WhatsApp: chat.whatsapp....
👉 Facebook: / circuitdigest
👉 Twitter: / circuitdigest
👉 Pinterest: / circuitdigest
👉 LinkedIn: / circuit-digest
-------------------------------------------------------------------------------------------------------------------------
Glad to see your video made 10x my views in just 10 days T_T. That said, I'm happy that more people get to know my library
Hey thank you! your work is also amazing, great job!!!
I think its possible, not sure though
@@Circuitdigest hi sir,
Which algorithm is this working on
Yours is the only Indian Tech Channel worth watching, all others make such basic videos, but your projects has some standard.
Thank you keep upgrading us
Nice video 👍 gonna try it asap. Keep uploading informative videos like this.
Thank you, I will
Great topic, thanks 👍
Thanks. Very nice job
very nice.. enjoyed
Gald to see Indian IT guy working also on hardware
Nice video, but zoom in on the Arduino IDE text so it's easier to view on a mobile device. Thanks!
You can get all the necessary details on the project page linked in the video description.
could you do a video with servos pan and tilt and esp32 cam (recognizing and tracking?)
Can you do a video on how to debug an Edge/YOLO project using a JTAG debugger (ESP-Prog)?
Thanks
For the ESP32 board, what version of the library is used?
2.0.4
Excellent video. Thank you. Few questions: 1) do you know if the libraries included in VScode? 2) Any idea how far can such a detection work with this camera?
You can use the same library with VScode and Platform IO. the detection range would be very small since the ESP32 can only process image with lower resolution. if you increase the resolution the frame rate will be reduced drastically and the ESP32 may even crash with very high resolutions.
Nice video❤
I want to on red led when onion is detected, green when potato is detected and yellow when tomato is detected, is it possible ? , if yes how can i do this ?
Yes you can activate a corresponding LED when a specific object is detected. You can check for the bb.label value and activate the LEDs accordingly.
nicely done! quick clarification - @5:39 GPIO0 should be held High or Low ?
GPIO should be pulled to LOW.
how to upgrade plan for edge impulse? I am not able to train the model for a large dataset
You will have to do that through the edge impulse website. The professional plan costs around 400$ per month.
Sir, can we use BLE 33 SENSE as programmer?
Effect of ambient light on performance??
Since we are working with gray scale images lighting difference is tolerable. But if you want reliable results lighting should be maintained constant
Building a High-Resolution POV Display using ESP3
I am making your this project in my third year
Sir Can you provide me your PCB garba file in this (please sir help)🙏🙏🙏
One more question is I can use the ESP 32 inbuilt data transfer and charging pin it is work or not you can tell me 🙏🙏
👇👇👇
Building a High-Resolution POV Display using ESP32
Gerber file already on our website
Sir, this is my first time on this channel. Sir, I made a car detection project using Edge Impulse, but when I uploaded it to the ESP32-CAM, I encountered many initializer errors during compilation. Sir, please tell me why this happened when everything was auto-generated.
You will have to give a bit more context.
Hi, I tried to follow and uploaded it and it has a problem c:\Users\USER!\Documents\Arduino\libraries\test_inferencing\src\edge-impulse-sdk\tensorflow\lite\micro\kernels\depthwise_conv.cpp:1842:80: error: either all initializer clauses should be designated or none of them should be
1842 | data_dims_t filter_dims = {.width = filter_width, .height = filter_height, 0, 0};
| ^
exit status 1
Try editing the file ei_classifier_config.h in exported Arduino library folder: /scr/edge-impulse-sdk/classifier/:
and set #define EI_CLASSIFIER_TFLITE_ENABLE_ESP_NN from 1 to 0
hi, may i know why when i uploaded the code it showed compilation error and also some In function 'TfLiteStatus tflite::{anonymous}::Prepare(TfLiteContext*, TfLiteNode*)': , either all initializer clauses should be designated or none of them should be
1789 | .channels = input->dims->data[3], 1
Make sure to follow steps exactly as in the article. If you still face the same issue try editing the file ei_classifier_config.h in exported Arduino library folder: /scr/edge-impulse-sdk/classifier/:
Disabling #define EI_CLASSIFIER_TFLITE_ENABLE_ESP_NN 1 and set it to 0 will allow the project to compile .
hello sir, I absolutely loved your video. I have a query, pls help me look into it. Ater saving impulse, when raw features are to be generated, it is saying "Error: VipsJpeg: Invalid SOS parameters for sequential JPEG" please help sir, this is causing a major setback in my projject.
Either change your training images. Images taken with some phones may show this kind of error. Use the ESP32 CAM to take the training images.
@@CircuitdigestThankyou, I did find out that this is caused by samsung phones, as pics taken using samsung devices have some error data, resaving the image in another device or format will work.
Can't upload camera code I get error 'Compilation error: 'ledcSetup' was not declared in this scope; did you mean 'ledc_stop'?
make sure to install Eloquent ESP32-CAM Library.
Bro after build the AI trained Model.the code is not Compiling it takes long time and they display "compilation terminated error". Please help me bro.to solve this problem.
Nobody can help you if you don't provide the exact error code.
@@Circuitdigest c:\Users\Mohamed Mushraf\Documents\Arduino\libraries\Waste_Segregation_System-project-1_inferencing\src\edge-impulse-sdk\porting\espressif\ESP-NN\src\convolution\esp_nn_conv_s8_filter_aligned_input_padded_esp32s3.S:2: fatal error: opening dependency file C:\Users\Mohamed Mushraf\AppData\Local\Temp\arduino\sketches\D60EA5CD2FB35EF678691B1E576D39DC\libraries\Waste_Segregation_System-project-1_inferencing\edge-impulse-sdk\porting\espressif\ESP-NN\src\convolution\esp_nn_conv_s8_filter_aligned_input_padded_esp32s3.S.d: No such file or directory
2 | #if EI_CLASSIFIER_TFLITE_ENABLE_ESP_NN && EI_CLASSIFIER_TFLITE_ENABLE_ESP_NN_S3
compilation terminated.
exit status 1
Compilation error: exit status 1
Why are you using ESP32 S3? which camera module are you using? For normal ESP32 camera module select the appropriate ESP32 board.
Will the accuracy with esp32 be lower or only slower compared to a solution on RP or PC? Which Yolo model is used for this ESP32 application?
The frame rate and accuracy will definitely better on RP or PC. We selected ESP32 cam because its a low cost option
@@Circuitdigest it is a great solution but I not understand why the same model would be less ac
curate on an esp32, slower I agree, but less accurate? Sure?
@@SA-oj3bo During training the model we reduced the training cycle so as to reduce the code size and make it easy for esp32. ideally if you need better accuracy you should provide more training datasets and use high epoch but this is not feasible on a small processor like esp32. also we cannot use a better resolution camera with esp32
it says error compiling for ai thinker
while uploading
You will have to be a bit more specific. what is the actual error?
How to get an output (when detect union.output high or low)
check from which variable the detection is coming from, and use if condition to check the class. Finally, use digitalWrite() to perform HIGH & LOW operations. Thanks!
You can use the GPIO pins on ESP32CAM to get output based on what you detect
Thank you for the video. I have a problem with the compiling of the code.
c:\Users\start\Documents\Arduino\libraries\Antress-project-1_inferencing\src\edge-impulse-sdk\tensorflow\lite\micro\kernels\conv.cpp: In function 'void tflite::{anonymous}::EvalQuantizedPerChannel(TfLiteContext*, TfLiteNode*, const TfLiteConvParams&, const NodeData&, const TfLiteEvalTensor*, const TfLiteEvalTensor*, const TfLiteEvalTensor*, TfLiteEvalTensor
Tka*)':
c:\Users\start\Documents\Arduino\libraries\Antress-project-1_inferencing\src\edge-impulse-sdk\tensorflow\lite\micro\kernels\conv.cpp:1881:58: error: either all initializer clauses should be designated or none of them should be
1881 | .channels = input_depth, 1
| ^
c:\Users\start\Documents\Arduino\libraries\Antress-project-1_inferencing\src\edge-impulse-sdk\tensorflow\lite\micro\kernels\conv.cpp:1885:59: error: either all initializer clauses should be designated or none of them should be
1885 | .channels = output_depth, 1
| ^
exit status 1
Compilation error: exit status 1
It's a piece of the error. I used esp 32 cams + Arduino ide 2.3.2 + edge impulse os was shown in the video but it did not work. Also, I visited the web page and recreated all steps, but without success. Which could be a problem?
Thank you for answer :)
What version of ESP32 Arduino core are you using? Make sure to follow steps exactly as in the article. If you still face the same issue try editing the file ei_classifier_config.h in exported Arduino library folder: /scr/edge-impulse-sdk/classifier/:
Disabling #define EI_CLASSIFIER_TFLITE_ENABLE_ESP_NN 1 and set it to 0 will allow the project to compile .
Sir how much project cost goes
Hardware wis you will need an ESP32-Cam module, which will cost anywhere around 400-600 INR.
Hello everyone...is this a supervised learning
yes
Yes this comes under supervised learning
are you a malayali ?
hello sir, is it not possible to have a bounding box or something similar? The point is to detect objects in real time with an ESP32 cam that has a bounding box or centeroid 😭😮💨
bro you don't start a video by calling the audience people