AI enabled acoustic object detection system
-
I'm developing an audio classification project with four classes using an RNN with LSTM. While the model performs well in classifying audio in silent environments, it struggles with noisy situations like crowds or disturbances. How can I train the model more robustly to accurately predict classifications in such challenging environments?
-
Hi @sanal143, Welcome to the MakerGram community
.Does your training data contain noise data similar to the data you are inferencing? and May I know, what kind of hardware you are using to run the model. Thanks
Recent Posts
-
Hi makers
i have choosen project theme.
project title : Still Heard
“For the ones who stayed quiet but always cared.”
Still Heard is an interactive emotional companion built for those who’ve ever felt unheard, unseen, or silenced especially in professional or personal spaces.
looked for some option.
need to take the input from the user and we have two options
direct text from user
or
capturing emotion using sensors
i have read about Grove-GSR sensor
can we use that one?
Is there any better options ? -
Hi @rahuljeyaraj ,
Welcome to the MakerGram community forum, There are several capable development boards available right now for the application. I'm sharing the one we have at our MakerGram inventory,
For small models, I recommend using - XIAO ESP32-S3 Sense along with Edge Impulse platform to create the model, Here the processing is managed by the esp32 Core itself. It maybe feel slower if the mode it bit heavy.
next option for small models is Seeed Grove Vision v2 camera where all the processing will done in the inbuilt "Arm Cortex-M55 & Ethos-U55" and you can get the inference results via the I2C or UART from the board, later you can process with Arduino or any XIAO board itself.
If the model is moderate in size, better to go with Raspberry Pi and v2 or v3 camera and use the Edge Impulse models.
Currently, we have these board available.
-
@salmanfaris Thank you Salman, This solution worked for me.
-
@Abhay Can you try one more method
Downloaded both boards, “Seeed nRF52 Boards” and “Seeed nRF52 mbed-enabled Boards.”
Copy the adafruit-nrfutil from
/Users/<user>/Library/Arduino15/packages/Seeeduino/hardware/mbed/2.9.2/tools/adafruit-nrfutil/macos/adafruit-nrfutilto
/Users/<user>/Library/Arduino15/packages/Seeeduino/hardware/nrf52/1.1.8/tools/adafruit-nrfutil/macos/adafruit-nrfutilThen chmod -x the copied file
Please check and let me know. Src