Error while trying to run model on Raspberry pi4

I have downloaded the Linux (AARCH64) file after the model was built and Im using as a guide, the code in the github repository to deploy my model in the Raspberry Pi4 with the Pi camera module3. However, i got the following error

I have installed all the dependecies:
"pip3 install edge_impulse_linux -i" &
sudo apt-get install libatlas-base-dev libportaudio0 libportaudio2 libportaudiocpp0 portaudio19-dev
Additionally, im not using the camera legacy mode.

Project ID: 216037

Context/Use case:

Currently the code looks like this:

#!/usr/bin/env python
Pi Camera Live Image Classification

Detects objects in continuous stream of images from Pi Camera. Use Edge Impulse
Runner and downloaded .eim model file to perform inference. Bounding box info is
drawn on top of detected objects along with framerate (FPS) in top-left corner.

Author: EdgeImpulse, Inc.
Date: August 3, 2021
Updated: August 9, 2023
License: Apache-2.0 (

import os, sys, time
import cv2
from picamera2 import Picamera2
from edge_impulse_linux.image import ImageImpulseRunner


model_file = “ml-part-quality-160x160_v1-linux-aarch64-v15.eim” # Trained ML model from Edge Impulse
res_width = 160 # Resolution of camera (width)
res_height = 160 # Resolution of camera (height)
rotation = 0 # Camera rotation (0, 90, 180, or 270)

The ImpulseRunner module will attempt to load files relative to its location, so we make it load files relative to this program instead

dir_path = os.path.dirname(os.path.realpath(file))
model_path = os.path.join(dir_path, model_file)

Load the model file

runner = ImageImpulseRunner(model_path)

Initialize model (and print information if it loads)

model_info = runner.init()
print(“Model name:”, model_info[‘project’][‘name’])
print(“Model owner:”, model_info[‘project’][‘owner’])

Exit if we cannot initialize the model

except Exception as e:
print(“ERROR: Could not initialize model”)
print(“Exception:”, e)
if (runner):

Initial framerate value

fps = 0

Interface with camera

with Picamera2() as camera:

# Configure camera settings
config = camera.create_video_configuration(main={"size": (res_width, res_height)})

# Start camera

# Continuously capture frames
while True:
    # Get timestamp for calculating actual framerate
    timestamp = cv2.getTickCount()
    # Get array that represents the image (in RGB format)
    img = camera.capture_array()

    # Rotate image
    if rotation == 0:
    elif rotation == 90:
        img = cv2.rotate(img, cv2.ROTATE_90_CLOCKWISE)
    elif rotation == 180:
        img = cv2.rotate(img, cv2.ROTATE_180)
    elif rotation == 270:
        img = cv2.rotate(img, cv2.ROTATE_90_COUNTERCLOCKWISE)
        print("ERROR: rotation not supported. Must be 0, 90, 180, or 270.")
    # Extract features (e.g. grayscale image as a 2D array)
    features, cropped = runner.get_features_from_image(img)
    # Perform inference
    res = None
        res = runner.classify(features)
    except Exception as e:
        print("ERROR: Could not perform inference")
        print("Exception:", e)
    # Display predictions and timing data
    results = res['result']['classification']
    for label in results:
        prob = results[label]
        print(label + ": " + str(round(prob, 3)))
    print("FPS: " + str(round(fps, 3)))
    # Find label with the highest probability
    max_label = max(results, key=results.get)

    # For viewing, convert image to grayscale
    img = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
    # Draw max label on preview window
                (0, 12),
                (255, 255, 255))
    # Draw max probability on preview window
                str(round(results[max_label], 2)),
                (0, 24),
                (255, 255, 255))
    # Show the frame
    cv2.imshow("Frame", img)
    # Calculate framrate
    frame_time = (cv2.getTickCount() - timestamp) / cv2.getTickFrequency()
    fps = 1 / frame_time
    # Press 'q' to quit
    if cv2.waitKey(1) == ord('q'):

Clean up


Am I missing something? I would really appreciate it if I could get some help.

@shawn_edgeimpulse should be able to answer this.

Maybe you trained a grayscale model and Shawn’s code wants a RGB.

Hello @MMarcial

No, I have a trained the model with RGB images. :smiling_face_with_tear:

Hi @victorinox,

Can you verify that your PiCam 3 works with the preview code here:

Hello @shawn_edgeimpulse

Yes its working! Although the image quality its a bit low is it linked with the 320 by 320 height and width set in the code?
picam works

Hi @victorinox,

It looks like that error is from a bug in my code. It appears the picamera2 library is defaulting to a 4-channel model (RGB8888) that includes the alpha channel, and get_features_from_image() expects 3 channels (RGB888). To fix that, change the camera config to the following:

config = camera.create_video_configuration(
        main={"size": (res_width, res_height), "format": "RGB888"})

I have updated the code in my course examples To include that change.

The quality looks low because your camera is out of focus. Out of the box, the Pi Camera has the focus set to infinity (or at least a few meters away). You can adjust the focus by using rotating the plastic housing around the lens (see here). I recommend doing this carefully with some needle-nose pliers.


Hello @shawn_edgeimpulse!

The model is working! Thank you very much :grin:

However, the image is in grayscale is there a way to put in RGB?

Edit: I have found the line in the code which converts RGB to Grayscale. I just Hashtagged it.

    # For viewing, convert image to grayscale
    img = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)