Chapter 3. application solution
3.2.5 What model frameworks does ESP-DL support?
Currently, ESP-DL supports models from mxnet, pytorch, and tensorow.
3.2.6 Does ESP-DL support all models of the three platforms mentioned above (mxnet,
pytorch, and tensorow)?
ESP-DL supports models in which all the operators are supported by ESP-DL. Please check layer for
the supported operators.
3.2.7 Can the model les of ESP-SKAINET be stored in the SD card?
Yes.
3.2.8 How to customize command words in ESP-SKAINET?
To customize command words, please see ESP-SR GitHub.
3.2.9 How to reduce the system footprint of AI speech models?
You can choose to turn o the three functions, namely AEC, AE, and VAD.
3.2.10 What is the dierence between a 16-bit quantization model and an 8-bit quanti-
zation model?
The 16-bit quantization model has higher precision and more accurate results, while the 8-bit quantization
model is more lightweight.
3.2.11 How does the AI voice model modify the number of microphone channels?
The number of microphone channels and the number of playback channels can be congured in the
AFE.
3.2.12 How do I get the actual audio captured in the development board?
To obtain the actual audio, an SD card interface is required to store the audio les to the SD card.
Espressif Systems 24
Submit Document Feedback
Release master