Import Cv2 error can't solve by adding opencv headless to requirments.txt

[Mr_Algorithms]

the app gives an eroro in deployment i read the other topics but it’s still giving me the same error

[Mr_Algorithms]

this is my repo

GitHub - marioeid/Facial-Emotion-recognition-streamlit: Detecting the...

[Franky1]

Error while deploying an app with mediapipe

@ajaygithub2 Try this one: requirements.txt streamlit mediapipe packages.txt libgl1-mesa-glx libglib2.0-0 test.py import streamlit as st import cv2 import mediapipe as mp print(st.__version__) print(cv2.__version__) print(mp.__version__) gave me the folowing log output in a Python 3.9. environment 1.21.0 4.7.0 0.9.2.1 The above quick solution should work. IMHO the problem with mediapipe is, that it enforces the installation of a non-headless version of opencv, which causes issues in a…

[Mr_Algorithms]

streamlit is 1.26.0
cv2 is 4.8.0
media pipe is 0.9.0.1

[Mr_Algorithms]

streamlit is 1.26.0
cv2 is 4.8.0
media pipe is 0.9.0.1

[Franky1]

requirements.txt

numpy
pandas
streamlit
streamlit_webrtc==0.47.0
av==10.0.0
deepface==0.0.75
mediapipe==0.9.0.1
tensorflow-cpu==2.13.0

Do not add opencv here!

packages.txt

libgl1-mesa-glx
libglib2.0-0

[Mr_Algorithms]

well i tried something else and it worked i just added packagest.txt and it worked fine
this is the app
https://facial-emotion-recognition-app-mralgorithms.streamlit.app/
thank you alot

[Mr_Algorithms]

open cv is still added but only worked after i added the packages.txt file no idea why

[Franky1]

@Mr_Algorithms

only worked after i added the packages.txt file no idea why

Because mediapipe unfortunatey enforces the installation of the non-headless opencv version and this opencv version requires some binaries preinstalled. Therefore you need packages.txt and you can remove opencv from the requirements.txt file. Also remove tensorflow_intel because this may not work on streamlit cloud.