Deploy custom YOLOv8 model to OAK D-Pro W

HassanGeedi

I'm having trouble deploying my custom YOLOv8 model to my oak camera with spatial detectio. It keeps on throwing a [error] Not enough memory to start a given neural network, needed at least: 38999040B

If i pass spatial=Falsethe script keeps running without any results

How can i fix this ?

Here is the full error:

Here is my script:

import os

import json

from depthai_sdk import OakCamera

HOME = os.getcwd()

with OakCamera(usb_speed="usb2") as oak:

color = oak.create_camera('color')

nn = oak.create_nn(f"{HOME}/result/latest_openvino_2022.1_4shave.blob",

color, nn_type='yolo', spatial=True, tracker=True)

oak.visualize(nn, fps=True, scale=2/3)

file = open(f"{HOME}/result/latest.json")

conf = json.load(file)

nn.config_yolo_from_metadata(conf['nn_config']['NN_specific_metadata'])

oak.visualize(nn.out.passthrough, fps=True)

oak.start(blocking=True)

jakaskerl

Hi HassanGeedi
Most likely it's the custom model problem. Perhaps there is an operation/layer that is too computationally expensive. Or maybe try changing the openvino version.

Hope this helps,
Jaka