The toolkit to generate auto labeled datasets

Last update: Mar 28, 2022

Overview

Ozeu

Ozeu is the toolkit to autolabal dataset for instance segmentation. You can generate datasets labaled with segmentation mask and bounding box from the recorded video files.

Installation

Requirements

ffmpeg
torch
mmcv-full

Example installation command for cuda11.1.

pip install torch==1.8.1+cu111 torchvision==0.9.1+cu111 torchaudio==0.8.1 -f https://download.pytorch.org/whl/torch_stable.html

pip install mmcv-full==1.3.5 -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/index.html

pip install git+https://github.com/open-mmlab/[email protected]

git clone [email protected]:xiong-jie-y/ozeu.git
cd ozeu
pip install -e .

Usage

1. Record Video

I recommend record video with the camera where you want to run detector. For webcam, you can use command like this.

ffmpeg -f v4l2 -framerate 60 -video_size 1280x720 -i /dev/video0 output_file.mkv

I recommend to place the object to record in a desk or somewhere on simple texture. That will reduce error rate. You can hold the object by your hand, because the dataset generator can recognize and remove hand like this.

2. Create dataset definition file.

You can write dataset definition file in yaml. Please define class names and ids at categories, and please associate class id and video paths in the datasets. The class ids will be the label of the files. video_path is relative to the dataset definition file. Video files that are supported by ffmpeg can be used.

categories:
  - id: 1
    name: alchol sheet
  - id: 2
    name: ipad
datasets:
  - category_id: 2
    video_path: IMG_4194_2.MOV
  - category_id: 2
    video_path: IMG_4195_2.MOV

3. Generate labaled coco dataset.

You can generate labaled coco dataset by giving the dataset definition file above. If you didn't hold object by hand while recording video, you can remove --remove-hand option.

python scripts/create_coco_dataset_from_videos.py  --dataset-definition-file ${DATASET_DEFINITION_FILE} --model-name u2net --output-path ${OUTPUT_DATASET_FOLDER} --resize-factor 2 --fps 15 --remove-hand

4. Generate background augmented datasets.

Please place background images at backgrounds_for_augmentation. The background augmentation script will use these files to replace background of datasets. Here we use VOC images as background images

wget https://pjreddie.com/media/files/VOCtrainval_11-May-2012.tar
--2021-06-02 22:13:22--  https://pjreddie.com/media/files/VOCtrainval_11-May-2012.tar
tar xf VOCtrainval_11-May-2012.tar
mkdir backgrounds_for_augmentation
mv VOCdevkit/VOC2012/JPEGImages/* backgrounds_for_augmentation/

After preparing background images, please generate background augmented dataset by running

python scripts/generate_background_augmented_dataset.py --input-dataset-path ${DATASET_FOLDER} --destination-root ${AUGMENTED_DATASET_FOLDER} --augmentation-mode different_background

5. Merge

You can merge background augmented dataset and dataset.

python scripts/merge_coco_datasets.py --input-dirs ${AUGMENTED_DATASET_FOLDER} --input-dirs ${DATASET_FOLDER} --destination-root ${MERGED_DATASET}

6. (Optional) Import dataset into cvat.

There is the annotation tool CVAT that can accept coco format dataset. So you can import dataset into your project and fix dataset.

7. TRAIN!

TRAIN!!!

Acknowledgement

I wish to thank my wife, Remilia Scarlet.
This toolkit uses U^2 net for salient object detection. Thank you for nice model!

Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit

CNTK Chat Windows build status Linux build status The Microsoft Cognitive Toolkit (https://cntk.ai) is a unified deep learning toolkit that describes

17k Feb 11, 2021

Official PyTorch implementation of "Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets" (ICLR 2021)

Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets This is the official PyTorch implementation for the paper Rapid Neural A

48 Dec 26, 2022

A data annotation pipeline to generate high-quality, large-scale speech datasets with machine pre-labeling and fully manual auditing.

About This repository provides data and code for the paper: Scalable Data Annotation Pipeline for High-Quality Large Speech Datasets Development (subm

86 Dec 7, 2022

Asterisk is a framework to generate high-quality training datasets at scale

44 Apr 25, 2022

Experimenting with computer vision techniques to generate annotated image datasets from gameplay recordings automatically.

Experimenting with computer vision techniques to generate annotated image datasets from gameplay recordings automatically. The collected data will then be used to train a deep neural network that can detect enemy player models in real time, during gameplay. Finally, a virtual input device will adjust the player's crosshair based on live detections for greater accuracy.

3 Apr 24, 2022

根据midi文件演奏“风物之诗琴”的脚本 "Windsong Lyre" auto play

Genshin-lyre-auto-play 简体中文 | English 简介根据midi文件演奏“风物之诗琴”的脚本。由Python驱动,在此承诺， ⚠️ 项目内绝不含任何能够引起安全问题的代码。前排提示：所有键盘在动但是原神没反应的都是因为没有管理员权限，双击run.bat或者以管理员模式

386 Jan 1, 2023

Official implementation for Likelihood Regret: An Out-of-Distribution Detection Score For Variational Auto-encoder at NeurIPS 2020

Likelihood-Regret Official implementation of Likelihood Regret: An Out-of-Distribution Detection Score For Variational Auto-encoder at NeurIPS 2020. T

33 Oct 12, 2022

Add-on for importing and auto setup of character creator 3 character exports.

CC3 Blender Tools An add-on for importing and automatically setting up materials for Character Creator 3 character exports. Using Blender in the Chara

260 Jan 5, 2023

Offcial repository for the IEEE ICRA 2021 paper Auto-Tuned Sim-to-Real Transfer.

47 Jun 30, 2022

Releases(0.0.1dev4)

0.0.1dev4(Jun 22, 2021)
Remove unnecessary IPython.embed()

Source code(tar.gz)
Source code(zip)
0.0.1dev3(Jun 22, 2021)
Bug Fix: Fix the bug in which video file is skipped when creating coco dataset.

Source code(tar.gz)
Source code(zip)
0.0.1dev2(Jun 3, 2021)
Updated installation procedure to the correct one.

Source code(tar.gz)
Source code(zip)
0.0.1dev1(Jun 2, 2021)
Feature to autolabel labaled coco dataset from videos.

Procedures to create dataset.

Feature to remove hand.

Source code(tar.gz)
Source code(zip)

The toolkit to generate auto labeled datasets

Related tags

Overview

Ozeu

Installation

Requirements

Example installation command for cuda11.1.

Usage

1. Record Video

2. Create dataset definition file.

3. Generate labaled coco dataset.

4. Generate background augmented datasets.

5. Merge

6. (Optional) Import dataset into cvat.

7. TRAIN!

Acknowledgement

You might also like...

Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit

Official PyTorch implementation of "Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets" (ICLR 2021)

A data annotation pipeline to generate high-quality, large-scale speech datasets with machine pre-labeling and fully manual auditing.

Asterisk is a framework to generate high-quality training datasets at scale

Experimenting with computer vision techniques to generate annotated image datasets from gameplay recordings automatically.

根据midi文件演奏“风物之诗琴”的脚本 "Windsong Lyre" auto play

Official implementation for Likelihood Regret: An Out-of-Distribution Detection Score For Variational Auto-encoder at NeurIPS 2020

Add-on for importing and auto setup of character creator 3 character exports.

Offcial repository for the IEEE ICRA 2021 paper Auto-Tuned Sim-to-Real Transfer.

Releases(0.0.1dev4)

0.0.1dev4(Jun 22, 2021)

0.0.1dev3(Jun 22, 2021)

0.0.1dev2(Jun 3, 2021)

0.0.1dev1(Jun 2, 2021)

Owner

Xiong Jie

DimReductionClustering - Dimensionality Reduction + Clustering + Unsupervised Score Metrics

TransReID: Transformer-based Object Re-Identification

Hypercomplex Neural Networks with PyTorch

LAMDA: Label Matching Deep Domain Adaptation

FLSim a flexible, standalone library written in PyTorch that simulates FL settings with a minimal, easy-to-use API

Densely Connected Convolutional Networks, In CVPR 2017 (Best Paper Award).

Medical Insurance Cost Prediction using Machine earning

Structural Constraints on Information Content in Human Brain States

This repository contains Prior-RObust Bayesian Optimization (PROBO) as introduced in our paper "Accounting for Gaussian Process Imprecision in Bayesian Optimization"

Anomaly detection in multi-agent trajectories: Code for training, evaluation and the OpenAI highway simulation.

Code for our paper A Transformer-Based Feature Segmentation and Region Alignment Method For UAV-View Geo-Localization,

CFC-Net: A Critical Feature Capturing Network for Arbitrary-Oriented Object Detection in Remote Sensing Images

Physics-informed Neural Operator for Learning Partial Differential Equation

Code base for NeurIPS 2021 publication titled Kernel Functional Optimisation (KFO)

Julia package for contraction of tensor networks, based on the sweep line algorithm outlined in the paper General tensor network decoding of 2D Pauli codes

Code accompanying the paper Shared Independent Component Analysis for Multi-subject Neuroimaging

Hybrid Neural Fusion for Full-frame Video Stabilization

Code for "LASR: Learning Articulated Shape Reconstruction from a Monocular Video". CVPR 2021.

Predict bus arrival time using VertexAI and Nvidia's Jetson Nano

Deep Learning Specialization by Andrew Ng, deeplearning.ai.