Giter Club home page Giter Club logo

zero-shot-object-navigation's Introduction

Zero-Shot Object Goal Visual Navigation

This implementation is modeified based on MJOLNIR and SAVN.

The code has been implemented and tested on Ubuntu 18.04, python 3.6, PyTorch 0.6 and CUDA 10.1

Setup

  1. (Recommended) Create a virtual environment using virtualenv or conda:
virtualenv ZSON --python=python3.6
source ZSON/bin/activate
conda create -n ZSON python=3.6
conda activate ZSON
  1. Clone the repository as:
git clone https://github.com/pioneer-innovation/Zero-Shot-Object-Navigation.git
cd Zero-Shot-Object-Navigation
  1. For the rest of dependencies, please run
pip install -r requirements.txt --ignore-installed

Data

The offline data can be found here.

"data.zip" (~5 GB) contains everything needed for evalution. Please unzip it and put it into the Zero-Shot-Object-Navigation folder.

For training, please also download "train.zip" (~9 GB), and put all "Floorplan" folders into ./data/thor_v1_offline_data

Evaluation

Note: if you are not using gpu, you can remove the argument --gpu-ids 0

Evaluate our model under 18/4 class split

python main.py --eval \
    --test_or_val test \
    --episode_type TestValEpisode \
    --load_model pretrained_models/SelfAttention_test_18_4.dat \
    --model SelfAttention_test \
    --gpu-ids 0 \
    --zsd 1 \
    --split 18/4

Evaluate our model under 14/8 class split

python main.py --eval \
    --test_or_val test \
    --episode_type TestValEpisode \
    --load_model pretrained_models/SelfAttention_test_14_8.dat \
    --model SelfAttention_test \
    --gpu-ids 0 \
    --zsd 1 \
    --split 14/8

Training

Note: the folder to save trained model should be set up before training.

Train our model under 18/4 class split

python main.py \
    --title mjolnir_train \
    --model SelfAttention_test \
    --gpu-ids 0 \
    --workers 8 \
    --vis False \
    --save-model-dir trained_models/SA_18_4/ \
    --zsd 1 \
    --partial_reward 1 \
    --split 18/4

Train our model under 14/8 class split

python main.py \
    --title mjolnir_train \
    --model SelfAttention_test \
    --gpu-ids 0 \
    --workers 8 \
    --vis False \
    --save-model-dir trained_models/SA_14_8/ \
    --zsd 1 \
    --partial_reward 1 \
    --split 14/8

zero-shot-object-navigation's People

Contributors

pioneer-innovation avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar

zero-shot-object-navigation's Issues

Issue about similarity calculation

hello!
I‘d like to ask which part of the code performs the similarity matching function and how you obtain and deal with the sematic tags? I couldn't find the code about that in the given project. Could you please point out for me?
Thank you a lot!

error occurred

error occurred when run the main.py for evaluation,
I used this command:
image

image

The number of epoch?

May I ask how many training epoch of the pretrained models? I trained 900,000 epoch using your code, 18/4 success rate difference 0.06. The difference in spl is 0.02. What parameters do I need to modify?

About the evaluation process

Hi, I want to ask how the current frame of objects are obtained in evaluation process.
I would appreciate if you give me a reply.

error when training

File "main.py", line 225, in
main()
File "main.py", line 206, in main
f = open(save_path, "a")
FileNotFoundError: [Errno 2] No such file or directory: '/trained_models/SA_18_4/SelfAttention_test_runs/mjolnir_train-2024-04-08_18-41-46.txt'

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.