jaseemck/computervision-with-pytorch-learning-program

Computer Vision using PyTorch Learning Program by TinkerHub Foundation

ComputerVision with PyTorch Learning Program

Computer Vision using PyTorch Learning Program by TinkerHub Foundation. PyTorch is an open source deep learning framework created by Facebook AI research. This learning program will cover the following,

• Computer vision.
• Pytorch framework.
• Torchvision library.
• Image classification and object detection.
• Transfer learning.

We will be extensively using the PyTorch docs for conducting this program.

Participants criteria

• Should know object oriented programming and python.
• Should know Git and GitHub.
• Should know what is machine learning and some basics(different categories of ML, what is training ? What is testing ? What is dataset..etc)

All the resources to get you started with the program is given in the resources folder. You can learn it and finish the task for joining the program!

Join the program

This learning program need you to have knowledge in following areas,

1. Python
2. GitHub
3. Basics of Machine Learning

We have a coding task for selction process. You could check out the task and how to submit task here. Watch the following video for reference.

Selection process

• once the task is completed, You could fill this google form

• we will update the selction status latest by 5th June.

• Program starts on 6th June.

General Program Structure

Every 2 days we will share a chapter of the PyTorch docs in the morning. Participants should go through it. In the evening of the odd days we will give a task. On the following day there will be hangouts sessions with mentors.

Program Schedule

Day 1

Google meets with the following contents:

• Welcome note to TinkerHub Computer vision using PyTorch program.
• Intro to PyTorch, navigate through PyTorch website.
• Reading documentations. Explaining the structure of PyTorch docs.
• General structure of the Program.
• Prerequisites, installations, getting to know each other.

Day 2

Morning : In the telegram channel we share the following content.

1. What is PyTorch ?
2. Autograd: Automatic Differentiation

Participants should,

• Go through the docs and learn.
• Try out each of the code snippets in the docs.
• Note down the doubts.
• Ask doubts in the github repo issues.

Evening : A task will be given in the telegram channel. Participants should try to do the task, upload the code to the specified GitHub repo.

Day 3

Morning : In the telegram channel we share the following content.

1. Neural Networks
2. Training a Classifier

Participants should,

• Go through the docs and learn.
• Try out each of the code snippets in the docs.
• Note down the doubts.
• Ask doubts in the github repo issues.

Evening : A task will be given in the telegram channel. Participants should try to do the task, upload the code to the specified GitHub repo.

Day 4

Google meets sessions with the mentor in the evening.

• Doubts about the last 2 days topics can be asked.
• Solution of the task will be discussed.

Day 5

Morning : In the telegram channel we share the following content.

1. What is torch.nn really ?
2. Visualising models, data and training with TensorBoard

Participants should,

• Go through the docs and learn.
• Try out each of the code snippets in the docs.
• Note down the doubts.
• Ask doubts in the github repo issues.

Evening : A task will be given in the telegram channel. Participants should try to do the task, upload the code to the specified GitHub repo.

Day 6

Google meets session with mentor(s) in the evening,

• Explain the concepts of content shared last day.
• Explain the code line by line.
• Clear doubts.

Day 7

Google meets session on,

1. What is computer vision ?
2. Algorithms used for computer vision?
3. TorchVision library.

Day 8

In the morning the following content will be shared in telegram.

1. TorchVision fine tuning for computer vision tutorial
2. Transfer learning for computer vision tutorial

Participants should,

• Go through the doc and try to understand.
• Note down the doubts.
• Try out the code.
• Ask doubts in the github repo issues.

Day 9

Google meets sessions with the mentor in the evening.

• Doubts about the last days topics can be asked.
• Solution of the task will be discussed.

Day 10

Google meets session with mentor(s) on,

• Choosing projects. Participants can choose transfer-learning projects or object detection projects.
• Finding a dataset. Can find a dataset from kaggle..etc.

Day 11

Participants figuring out the project and dataset. Participants should,

• Create a GitHub repo for the project.
• Update the Readme file with the details of the project.
• Find the model and dataset they are going to implement.
• Fill the project form with the project repo link.

Day 12

Mentor(s) verify the project ideas. Participants make the changes.

Day 13-14

Project days.

• Participants do the project.
• Ask doubts to the mentors in github issues and via call.
• Upload the code and trained models to the repo.

Day 15

Mentors verify the projects. Provide feedback as GitHub issues.

Day 16

Final google meets session with,

• Discussion on next steps from here.
• Project demos.
• Certificates distribution.

Program partner

Facebook Developer Circle Kochi is a forum for developers in Kochi, India and its surroundings who are interested in building on the Facebook platform to interact and collaborate other developers who share similar interests.

Contributors

• GopikrishnanSasikumar