/Learning-Continuous-Implicit-Representation-for-Near-Periodic-Patterns

Official code for paper 'Learning Continuous Implicit Representation for Near-Periodic Patterns' (ECCV 2022)

Primary LanguagePython

Learning to Transform High-frequency Periodic Distributions with Guided Control

Get started

You can set up the environment with all dependencies like so:

conda create --name NPP-Net python=3.8.5
conda activate NPP-Net
pip install -r requirements.txt

High-Level structure

  • data: input examples for completion, remapping, and segmentation.
  • externel_lib: externel library to support our code.
  • loaders：dataloader
  • models: helper functions for model
  • options: arguments for training.
  • NPP_proposal: implementation for top-K periodicity proposal.
  • NPP_completion: implementation for completion task.
  • NPP_segmentation: implementation for segmentation task.
  • NPP_remapping: implementation for remapping task.

How to Run

  1. Please download the file (https://github.com/42x00/p3i) download the pre-trained AlexNet weight in the "Pre-trained Models" section.

  2. Put the downloaded file (alexnet-owt-4df8aa71.pth) in the root of this directory.

NPP Completion

Run all examples in the "data/completion/input" using the following command.

bash run_completion.sh

This script first searches the periodicity of the image, saved in "data/completion/detected". Then it performs image completion, generating the outputs in "results/completion_top3".

NPP Segmentation

Run all examples in the "data/segmentation/input" using the following command.

bash run_segmentation.sh

This script first searches the periodicity of the image, saved in "data/segmentation/detected". Then it performs image segmentation, generating the outputs in "results/segmentation_top3".

NPP Remapping

Run all examples in the "data/remapping/input" using the following command.

bash run_remapping.sh

This script first searches the periodicity of the image, saved in "data/remapping/detected". Then it performs image remapping, generating the outputs in "results/remapping_top3".

Disclaimer

The result produced by this code might be slightly different when running on a different GPU.

Extra details

This repository downloads VGG and AlexNet. Make sure you change your pip cache path when running this codebase on a remote server.

Change cache path for the session as follows: export PIP_CACHE_DIR=/workspace/shubham/local_cache export PIP_CACHE_DIR=/workspace/ashwin/local_cache export PIP_CACHE_DIR=/workspace/your_local_folder/local_cache