WEEK 3
Date: 14 February 2020
Date: 14 February 2020
Place: EE Building, 3th floor/ Room 304, 2P22
Before lab
l Finish installing mjpg-streamer which is for
the remote monitoring module and finish constructing the webpage.
l Learning the way to achieve remote
controlling function based on bottle.
l After completing automatic mode
contemplated portion, we would use Tensorflow lite to achieve the object
recognition function, which is based on Opencv. The program language we used was
python3.
During lab
l We tested the monitoring function based on
mjpg-streamer and it worked well. We also improved quality of images and
fluency, and our web design is simple and pretty.
l Start working on automatic part. The first
step is to build the recognition system. We would use Tensorflow lite to
achieve this function. TF lite is an open source library for object
recognition, which is based on Opencva — library of programming
functions that can shows real-time computer vision.
Problem and solution
The main problem we
faced this week is installing Tensorflow issues. When we tried to run the program to open the web-camera, it showed:
ImportError:
No module named cv2
We
spent too much time working on solving this problem. We tried to reinstall all
the files and ran it in other environment but it did not work. Moreover, we thought
the reason was that we used wrong version of Python. Our python was updated to
the newest version 3.8.1, but the system ran python2 by default. We tried to
run the program on Python2 and it worked, and real-time version can be shown
successfully.
However, the problem is that Tensorflow do not support Python. In
this case, we had to delate Python2 and used the newest version. Because of
that, the other part of this project—the manual control part has to be
reinstalled. After upgrading to the newest version, the Opencv is available but
not the Tensorflow. We did a lot of search and tried serval different ways, then
we found out that the problem is that the Tensorflow only updated to Python3.7
while python just updated to 3.8 early this year. It means there is no such
kind of version of Tensorflow matching Python3.8. We downgraded the Python
version back to 3.7.3 after serval attempts and our program worked finally.
Comments
Post a Comment