Body measurements guidance system using computer vision and pose estimation
The solution built is a Python Flask application that processes images of through a custom mobilenet model that is adpated from the Pose Net model.
- Python Flask
- Tensorflow
- Anaconda (Windows)
- Bluetooth serial (Arduino BluNo)
- Html/CSS/JS
- Python >= 3.6.5
- GPU (Optional)
- PIP installed
-
Install Anaconda from here (Anaconda >= 5.3)
-
Install Swig from here here
-
Open Anaconda prompt:
Go with the mouse to the Windows Icon (lower left) and start typing "Anaconda". There should show up some matching entries. Select "Anaconda Prompt".
-
Clone our repository and change directory:
$ git clone https://github.com/Sri-vatsa/Unit $ cd Unit
-
Create & Activate virtual env:
$ conda create --prefix=unit python=3.6.5
Note: allow conda to install all relevant packages for python 3.6.5 by typing
y
when promptedList all available conda environments: (Optional)
$ conda info --envs
Activate conda environment for Unit:
$ conda activate PATH/TO/ENV
e.g. PATH/TO/ENV :
D:\Unit\unit
-
Install dependencies:
$ pip install -r requirements.txt
-
Make paf for pose estimation post processing:
a. Change directory to pafprocess:
$ cd tf_pose_estimation/tf_pose/pafprocess
b. Build the pafprocess module
$ swig -python -c++ pafprocess.i && python3 setup.py build_ext --inplace
-
change directory back to root directory:
$ cd ../../..
- Run backend:
$ python main.py
- Find your local ip address:
In the anaconda prompt, type:
$ ipconfig
IPv4 Address is your local ip address.Take note of this address.
E.g. 172.17.10.153
- Run the product demo in a browser
a. Open any browser (Chrome is preferred)
b. In the addressbar, type : https://YOUR_IP_ADDRESS:5000
E.g. https://172.17.10.153:5000
Note: If the browser alerts that the site is insecure, click on advanced settings -> continue anyway