Pytorch Summer Hackathon 2020 Information Page

Project name: PyVinci

Main packages deployed:

• pyvinci-segmentation
• pyvinci-client
• pyvinci-server

Main use cases:

• 1: Hashtag generator (Using a panoptic segmentation model)
• 2: (Not released): Generate new image from image labels (Using layer based sequencing image generator with GANs with the labels/masks created by the panoptic segmentation model)

Team members:

• Santiago Norena
• Hector Mejia
• Nicolas David
• Alejandro Martinez
• Sahivi Gonzalez

PyVinci Initial Concept

---

PyVinci Use Case Diagram

VERSION 2 (Areas highlighted in red were not released into production)

PyVinci System Architecture Diagram

Cross-function / Swimlane diagram

VERSION 2 (Version released for competition)

---

Deployment Demo:

Accessing the domain name https://www.pyvinci.com/ will take user to the main page. (No longer in production)

New user needs to register, then login.

Once logged in, user can see list of projects, or create a new one.

After uploading images in new project page, the user can click "Begin Modeling".

After the ML model generates labels for all images, they will show up below each image.

User can click "Home" to create a new project or see list of projects.

---

It was built with:

• numpy
• python
• pytorch
• detectron2
• cv2
• psycopq2
• sqlalchemy
• posgresql
• docker
• go / golang
• go-fiber
• nginx
• Amazon Web Services (AWS): EC2, S3
• npm
• react
• miragejs

Acknowledgements:

• https://github.com/0zgur0/Seq_Scene_Gen