Home page

Deconstructing Deep Learning + ╬┤eviations

Drop me an email | RSS feed link : Click | Source? Refer to the repository
Format : Date | Title
  TL; DR
✎: Post, ⚛ : Deviation, 📖: Book notes, 🚀: Spaceship

Total number of posts : 85

Search for something in the blog

View My GitHub Profile ; Interactive projects

Go to index

7 July 2020

Action Recognition part 2

Reading time : ~1 min

by Subhaditya Mukherjee

Continuing the action recognition project.

Custom path class

We first need to define a custom path class which will give us the paths of the original dataset, the preprocessed one and the pretrained weights. I know, we could probably do it without a class but I am reimplementing here so lets just go with it for now. The only new thing here is @staticmethod. This is python syntax for declaring a method which does not need a class to be instantiated and are not dependant on its state. Basically, they would return a static method for any function which is passed as its parameter.

class Path(object):
    def db_dir(database):
        root_dir = "/home/subhaditya/Desktop/Datasets/UCF101/"
        output_dir = "/home/subhaditya/Desktop/Datasets/UCF101pre/"
        return root_dir, output_dir
    def model_dir():
        return "/home/subhaditya/Desktop/Datasets/UCF101pre/Models/c3d-pre.pth"
Related posts:  Super resolution  Federated Learning  Taking Batchnorm For Granted  A murder mystery and Adversarial attack  Thank you and a rain check  Pruning  Documentation using Documenter.jl  Dataset Bindings  DCGAN  Differentiable Programming