Pytorch Dataloader for FewSOL
Project description
FewSOL-DataLoader
This repo hosts the PyTorch dataloader for FewSOL dataset.
Using package
First install the package using
pip install FewSOLDataLoader
Setup
Step-1. Download the FewSOL dataset from https://irvlutd.github.io/FewSOL/#data
- There are four splits of the FewSOL dataset:
real_objects
: This is a real single object image split. Each object was captured from 9 anglesreal_clutter
: This is a real clutter image split extracted from the OCIDsynthetic_objects
: This is a synthetic single object image split made with 3D google objects. Each object was captured from 9 anglesgoogle_clutter
: This is a synthetic clutter image split made with 3D google objects
- Note: The synthetic portion of the dataset is created using Google 3D Scanned Objects dataset.
Step-2. Pass the extracted dataset directory path into the dataloader as shown in the following example
Example Usage
from FewSOLDataLoader import load_fewsol_dataloader
# Define the root directory
ROOT_DIR = os.getcwd()
# Define the dataset root directory using the join_path function
DATASET_ROOT_DIR = os.path.join(ROOT_DIR, 'FewSOL', 'data')
# Log that the FewSOL dataloader is being loaded
print('Loading FewSOL dataloader')
test = load_fewsol_dataloader(DATASET_ROOT_DIR, split="real_objects")
# Generate a random index within the range of the dataloader's length
idx = random.randint(0, len(test) - 1)
# Retrieve data from the dataloader for the random index
image_data, semantic_data, bounding_data, label, questionnaire, file_name, poses = test[idx]
Data Formats
-
Image Data Shape
# n x q x w x h # n = Number of total images # q = 3 : Color slots for RGB # w = Width of the Image # h = Height of the image
-
Semantic Segmentation Shape
# n x m x w x h # n = Number of total images # m = Total number of objects in the current images # w = Width of the Image # h = Height of the image
-
Detection Bounds Shape
# n x m x r # n = Number of total images # m = Total number of objects in the current images # r = 4 : x, y, width, height
-
Pose Information
# n x m x 4 x 4 # n = Number of total images # m = Total number of objects in the current images
-
Label Output/Description Shape
# m = Total number of objects in the images
Licenses
All files are licensed under the MIT license except for the below two inside FewSOL-DataLoader/src/FewSOLDataLoader/
SingleRealPose.py
- licensed under the NVIDIA Source Code License - Non-commercial as found here.CocoFormatConverter.py
- licensed under the CC BY 4.0 LEGAL CODE as found here.
Bibtex
Please cite FewSOL if it helps your research:
@INPROCEEDINGS{padalunkal2023fewsol,
title={FewSOL: A Dataset for Few-Shot Object Learning in Robotic Environments},
author={P, Jishnu Jaykumar and Chao, Yu-Wei and Xiang, Yu},
booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
doi={10.1109/ICRA48891.2023.10161143},
pages={9140-9146},
year={2023}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
FewSOLDataLoader-0.0.5.tar.gz
(23.4 kB
view hashes)
Built Distribution
Close
Hashes for FewSOLDataLoader-0.0.5-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | fc56a4b573ef56cecf6dc4df9cd7f34c3f7b720f9576897dcf867ff0d095c88d |
|
MD5 | 8dff009872f525d7ff84544fb19914dd |
|
BLAKE2b-256 | 16fe7b154d69328aff2122ec275b0f813e7a42be1f574f84e7dda5b48d5e6c35 |