This is the official implementation of the paper "Object Propagation via Inter-Frame Attentions for Temporally Stable Video Instance Segmentation".

Last update: May 03, 2022

Related tags

Deep Learning ObjProp

Overview

ObjProp

Introduction

This is the official implementation of the paper "Object Propagation via Inter-Frame Attentions for Temporally Stable Video Instance Segmentation".

Installation

This repo is built using mmdetection. To install the dependencies, first clone the repository locally:

git clone https://github.com/anirudh-chakravarthy/objprop.git

Then, install PyTorch 1.1.0, torchvision 0.3.0, mmcv 0.2.12:

conda install pytorch==1.1.0 torchvision==0.3.0 -c pytorch
pip install mmcv==0.2.12

Then, install the CocoAPI for YouTube-VIS

conda install cython scipy
pip install git+https://github.com/youtubevos/cocoapi.git#"egg=pycocotools&subdirectory=PythonAPI"

Training

First, download and prepare the YouTube-VIS dataset using the following instructions.

To train ObjProp, run the following command:

python3 tools/train.py configs/masktrack_rcnn_r50_fpn_1x_youtubevos_objprop.py

In order to change the arguments such as dataset directory, learning rate, number of GPUs, etc, refer to the following configuration file configs/masktrack_rcnn_r50_fpn_1x_youtubevos_objprop.py.

Inference

To perform inference using ObjProp, run the following command:

python3 tools/test_video.py configs/masktrack_rcnn_r50_fpn_1x_youtubevos_objprop.py [MODEL_PATH] --out [OUTPUT_PATH.json] --eval segm

A JSON file with the inference results will be saved at OUTPUT_PATH.json. To evaluate the performance, submit the result file to the evaluation server.

License

ObjProp is released under the Apache 2.0 license.

Citation

@article{Chakravarthy2021ObjProp,
  author = {Anirudh S Chakravarthy and Won-Dong Jang and Zudi Lin and Donglai Wei and Song Bai and Hanspeter Pfister},  
  title = {Object Propagation via Inter-Frame Attentions for Temporally Stable Video Instance Segmentation},
  journal = {CoRR},
  volume = {abs/2111.07529},
  year = {2021},
  url = {https://arxiv.org/abs/2111.07529}
}

This is the official implementation of the paper "Object Propagation via Inter-Frame Attentions for Temporally Stable Video Instance Segmentation".

Related tags

Overview

ObjProp

Introduction

Installation

Training

Inference

License

Citation

Owner

Anirudh S Chakravarthy

Extract MNIST handwritten digits dataset binary file into bmp images

Unsupervised Representation Learning via Neural Activation Coding

Pre-trained model, code, and materials from the paper "Impact of Adversarial Examples on Deep Learning Models for Biomedical Image Segmentation" (MICCAI 2019).

Recommendation algorithms for large graphs

TResNet: High Performance GPU-Dedicated Architecture

TopFormer: Token Pyramid Transformer for Mobile Semantic Segmentation, CVPR2022

Implementation of Artificial Neural Network Algorithm

We have implemented shaDow-GNN as a general and powerful pipeline for graph representation learning. For more details, please find our paper titled Deep Graph Neural Networks with Shallow Subgraph Samplers, available on arXiv (https//arxiv.org/abs/2012.01380).

Source code for "FastBERT: a Self-distilling BERT with Adaptive Inference Time".

Information-Theoretic Multi-Objective Bayesian Optimization with Continuous Approximations

Joint Unsupervised Learning (JULE) of Deep Representations and Image Clusters.

Personalized Transfer of User Preferences for Cross-domain Recommendation (PTUPCDR)

Unimodal Face Classification with Multimodal Training

Unofficial PyTorch Implementation for HifiFace (https://arxiv.org/abs/2106.09965)

DeepConsensus uses gap-aware sequence transformers to correct errors in Pacific Biosciences (PacBio) Circular Consensus Sequencing (CCS) data.

Implementations of the algorithms in the paper Approximative Algorithms for Multi-Marginal Optimal Transport and Free-Support Wasserstein Barycenters

InvTorch: memory-efficient models with invertible functions

This is the unofficial code of Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes. which achieve state-of-the-art trade-off between accuracy and speed on cityscapes and camvid, without using inference acceleration and extra data

Public Models considered for emotion estimation from EEG

Very Deep Convolutional Networks for Large-Scale Image Recognition