- The labs 22 DS team's work can be found in the archived folder
- app contains our flaskapp along with the folders and files necessary to deploy to AWS.
- .ebextensions to increase max_file_body_size when deploying our model to Elastic Beanstalk. Without this increase the endpoint won't work with video.
- model contains the files necessary to build the model with the correct architecture, weights, and labels.
- remote-docker is for AWS deployment. The dockerfile we use that compiles ffmpeg from source takes to long to deploy normally.
- templates contains our index route for API testing purposes.
- testing_video.mp4 for quick API testing
- app.py contains the Flask class used to run the app and the API POST endpoint
- Dockerfile - necessary to get opencv and ffmpeg to work on AWS EB.
- HelperFunctions - Helper functions that get called in the /api route.
- model_test.py - Run this file to test current model real time via webcam.
- ModelFunctions - Functions to build the model, feed an image, and return a predicted letter.
- Procfile - Necessary for the Gunicorn engine & deployment
- Requirements.txt - The libraries used for our model & API endpoint.
- unit_tests - Testing suite. Run from the command line.
- Notebooks contains the notebook used to build our Yolo models.
- Weight and cfg files get saved to the drive
- Our most upto date model weights and cfg files for the individual lessons can be found in FlaskApp/model
Create and navigate to a working directory and run the following.
git clone https://github.com/Lambda-School-Labs/signlingo-ds.git
cd signlingo-ds/FlaskApp
Create a virtual environment.
pipenv shell --python 3.7
If the pip files are incorrectly installing the libraries into the environment, delete them and run the following
pipenv install -r requirements.txt
Run the following to initialize the TEMPVID and TEMPPICS folders
python app.py
Launching the flask app (NOTE: depending on the terminal, export can be substituted for set)
export FLASK_APP=app.py
flask run
(Documentation on how to run using Docker can be found in the Dockerfile)
- Contact any of us for the login information.
- https://ds.thesignlingo.com/api
- Expects a POST request with the following three inputs.
- 'video' of type file
- 'expected' of type string
- 'right-handed' of type number (1 for right hand)
#too large body error on AWS and/or CORS error from frontend -> Typically to fix this error, you're meant to add files in the .ebextensions folder as we have done, but it did not seem to work for us. As a result, we had to SSH into the instance directly from the EB CLI and modify the proper file ourselves. Keep in mind you have to do this after every deployment as it will reset itself. (Try to figure out the .ebextensions issue if you can) Steps to resolve:
- When you originally set up the EB CLI and initialize with eb init, you'll want to create a new keypair and use that to SSH into the instance.
- eb ssh
- sudo su (be VERY careful making any changes with this enabled, but I found it necessary to edit the files that needed editing)
- cd /etc/nginx
- nano nginx.conf
- Lines to add:
- Directly under
http{in the next line, indented to match the rest within it.client_max_body_size 50M - Under
location / {in the next line, addadd_header Access-Control-Allow-Origin *;- This is to resolve CORS issues with requests coming from the front end.
- Directly under
- Save and exit
- On the AWS console (which you can access with
eb console), click Actions and restart the app server. After this the changes should be active.