Skip to content

simCLR with downstream tasks on steering angle prediction

Notifications You must be signed in to change notification settings

poudel-bibek/Contrastive

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Data

The Honda Research institute data that we used in the project (pre-processed by NeurIPS paper) can be found here

Task 1: Robust Regressor

Standard and Robust Models

  • A Standard ResNet implementation code is obatined from here
  • Robust model code is obtained from here

simCLR Model

  1. To pre-train the model using simCLR: Run the Pretrain.ipynb Jupyter Notebook file
  2. To fine-tune the model: Go to DM_Project/simCLR/fine_tune/ directory and run
python3 main.py

Task 2: Denoising Autoencoder

In order to run the autoencoder code, some steps need to be completed first:

  1. Download the .pt files from this link: https://drive.google.com/drive/folders/1w7GfDKQ_wsClzP8fkrbpOe-Q4nRMjGaj?usp=sharing. These files go in the saved_models folder within the autoencoder folder.
  2. Download the data from this link: https://drive.google.com/drive/folders/1jzoYxJuIaClt9Q_tloYxK06tuAJrYAFi?usp=sharing. These files go in the data folder within the autoencoder folder.
  3. Now, you can either open the autoencoder folder within the simCLR folder with VS Code and run the main.py file or you can navigate to the folder through the terminal and run the following command:
python3 main.py
  1. This will run the code for the autoencoder portion of the project.

Testing trained models:

To test a trained model on validation sets, run

python3 fine_tune_model.py

About

simCLR with downstream tasks on steering angle prediction

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •