CODEWITHSUNDEEP
tensorflowesp32yoloquantization
gillo04
gillo04

Reputation: 98

Trying to quantize YOLOv11 in tensorflow, is this topology normal?

I'm trying to quantize the YOLO v11 model in tensorflow and get this as a result: enter image description here enter image description here The target should be int8. Is this normal behaviour? When running it with tflite micro on an esp32 I quicly run out of memory, even though I allocate 5 MB (the model is 3MB). Could my problem be tied to this wierd topology? Or are there any ways to mitigate my memory issues? I'm a total noob, so any help is appreciated!

Upvotes: 0

Views: 190

Answers (1)

Andres Guido
Andres Guido

Reputation: 1

I'm currently dealing with the same project!

The topology is normal. Try modifying the ESP32 partition to fix this memory issue. Here is some reference: https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-guides/partition-tables.html

I tried this partition, and it works well:

# Name,   Type, SubType, Offset,  Size, Flags
# Note: if you have increased the bootloader size, make sure to update the offsets to avoid overlap
nvs,      data, nvs,     0x9000,  0x6000,
phy_init, data, phy,     0xf000,  0x1000,
factory,  app,  factory, 0x10000, 0x400000, 
model,    data,  ,       0x410000, 0x300000,

Also check this issue in the esp-idf official repo: https://github.com/espressif/esp-idf/issues/12588

I have to say that i cannot make inferences yet, i am currently dealing with some micro op resolver issues, but the project can be built.

I hope this contribution helps you!

Upvotes: 0

Related Questions