The PyTorch implementation of DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision.

Last update: Oct 23, 2021

Related tags

Deep Learning SCOT

Overview

DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

The PyTorch implementation of DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision.

@article{lan2021discobox,
  title={DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision},
  author={Lan, Shiyi and Yu, Zhiding and Choy, Christopher and Radhakrishnan, Subhashree and Liu, Guilin and Zhu, Yuke and Davis, Larry S and Anandkumar, Anima},
  journal={arXiv preprint arXiv:2105.06464},
  year={2021}
}

[ Paper ]

Introduction

This repository is the implementation of evaluating DiscoBox on PF-Pascal dataset.

This implementation is based on SCOT

Installation

Conda environment settings

conda create -n scot python=3.6
conda activate scot

cat /usr/local/cuda/version.txt
conda install pytorch=1.4.0 torchvision cudatoolkit=10.0 -c pytorch (if CUDA 10) 
conda install pytorch=1.4.0 torchvision cudatoolkit=9.0 -c pytorch (if CUDA 9) 

conda install -c anaconda scikit-image
conda install -c anaconda pandas
conda install -c anaconda requests
pip install gluoncv-torch

Pretrained Weights

Download the pretrained weights from this link.

Evaluation

Results on PF-PASCAL with res101:

    python evaluate_map_CAM.py --dataset pfpascal --thres img --backbone resnet101 --hyperpixel '(2,22,24,25,27,28,29)' --sim OTGeo --exp1 1.0 --exp2 0.5 --eps 0.05 --gpu 0 --classmap 1 --split test --alpha 0.05
    python evaluate_map_CAM.py --dataset pfpascal --thres img --backbone resnet101 --hyperpixel '(2,22,24,25,27,28,29)' --sim OTGeo --exp1 1.0 --exp2 0.5 --eps 0.05 --gpu 0 --classmap 1 --split test --alpha 0.10
    python evaluate_map_CAM.py --dataset pfpascal --thres img --backbone resnet101 --hyperpixel '(2,22,24,25,27,28,29)' --sim OTGeo --exp1 1.0 --exp2 0.5 --eps 0.05 --gpu 0 --classmap 1 --split test --alpha 0.15

All results

Method	Backbone	[email protected]	[email protected]	[email protected]
SCOT	ResNet-101	63.2	85.4	92.8
DiscoBox	ResNet-101	62.7	85.6	93.5

Training

Please look at this link for more details

The PyTorch implementation of DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision.

Related tags

Overview

DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision

Introduction

Installation

Conda environment settings

Pretrained Weights

Evaluation

All results

Training

Owner

Shiyi Lan

Official repository for the NeurIPS 2021 paper Get Fooled for the Right Reason: Improving Adversarial Robustness through a Teacher-guided curriculum Learning Approach

Code for unmixing audio signals in four different stems "drums, bass, vocals, others". The code is adapted from "Jukebox: A Generative Model for Music"

Implicit Deep Adaptive Design (iDAD)

EncT5: Fine-tuning T5 Encoder for Non-autoregressive Tasks

Simulation-based inference for the Galactic Center Excess

This repo is the official implementation of "L2ight: Enabling On-Chip Learning for Optical Neural Networks via Efficient in-situ Subspace Optimization".

Official implementation for "Symbolic Learning to Optimize: Towards Interpretability and Scalability"

This project provides the proof of the uniqueness of the equilibrium and the global asymptotic stability.

A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning

PyTorch implementation of "Supervised Contrastive Learning" (and SimCLR incidentally)

This repository contains the PyTorch implementation of the paper STaCK: Sentence Ordering with Temporal Commonsense Knowledge appearing at EMNLP 2021.

Python scripts for performing object detection with the 1000 labels of the ImageNet dataset in ONNX.

GT China coal model

LightningFSL: Pytorch-Lightning implementations of Few-Shot Learning models.

The code written during my Bachelor Thesis "Classification of Human Whole-Body Motion using Hidden Markov Models".

Very large and sparse networks appear often in the wild and present unique algorithmic opportunities and challenges for the practitioner

Built a deep neural network (DNN) that functions as an end-to-end machine translation pipeline

DIR-GNN - Discovering Invariant Rationales for Graph Neural Networks

PyTorch original implementation of Cross-lingual Language Model Pretraining.

Individual Tree Crown classification on WorldView-2 Images using Autoencoder -- Group 9 Weak learners - Final Project (Machine Learning 2020 Course)