Dynamic Neural Representational Decoders for High-Resolution Semantic Segmentation

Last update: Sep 12, 2022

Related tags

Deep Learning NRD_decoder

Overview

Dynamic Neural Representational Decoders for High-Resolution Semantic Segmentation

Requirements

This repository needs mmsegmentation

Training

To train the model(s) in the paper, run this command:

python tools/train.py ./configs/NRD/ade20k/NRD_r101_512x512_164k_ade20k.py

The batch size is 16 in this work. Please change the 'samples_per_gpu' in configs/base/datasets/.. accordingly

Evaluation

To evaluate my model at single-scale inference, run:

python tools/eval.py ./configs/NRD/ade20k/NRD_r101_512x512_164k_ade20k.py  {path-to-checkpoint-file}   --eval mIoU

Pre-trained Models

Results

Our model achieves the following performance on :

[Semantic segmentation results]

Model name	datasets	mIoU	mIoU (ms)
NRD-r101	ade20k (val)	44.01	45.62
NRD-x101	ade20k (val)	44.34	46.35
NRD-r101	pascal-context(val)	52.31 (59 classes)	54.1 (59 classes)
NRD-r101	pascal-context(val)	47.5 (60 classes)	40.9 (60 classes)
NRD-r50	Cityscapes (val)	79.8	80.8
NRD-r101	Cityscapes (val)	80.7	82.0

Contributing

The code is mostly taken from mmsegmentation mmsegmentation is released under the Apache 2.0 license.

Dynamic Neural Representational Decoders for High-Resolution Semantic Segmentation

Related tags

Overview

Dynamic Neural Representational Decoders for High-Resolution Semantic Segmentation

Requirements

Training

Evaluation

Pre-trained Models

Results

[Semantic segmentation results]

Contributing

Owner

Adelaide Intelligent Machines (AIM) Group

Code for paper ECCV 2020 paper: Who Left the Dogs Out? 3D Animal Reconstruction with Expectation Maximization in the Loop.

Implementation of the ivis algorithm as described in the paper Structure-preserving visualisation of high dimensional single-cell datasets.

Pytorch implementation of MaskGIT: Masked Generative Image Transformer

Activating More Pixels in Image Super-Resolution Transformer

Source code, data, and evaluation details for “Cross-Lingual Citations in English Papers: A Large-Scale Analysis of Prevalence, Formation, and Ramifications”

This is an official implementation for "ResT: An Efficient Transformer for Visual Recognition".

Pretraining Representations For Data-Efficient Reinforcement Learning

OpenLT: An open-source project for long-tail classification

Perform zero-order Hankel Transform for an 1D array (float or real valued).

Personalized Federated Learning using Pytorch (pFedMe)

Digan - Official PyTorch implementation of Generating Videos with Dynamics-aware Implicit Generative Adversarial Networks

On the Limits of Pseudo Ground Truth in Visual Camera Re-Localization

GarmentNets: Category-Level Pose Estimation for Garments via Canonical Space Shape Completion

Vehicle speed detection with python

Multi-Task Learning as a Bargaining Game

Alignment Attention Fusion framework for Few-Shot Object Detection

HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis

Code for DeepCurrents: Learning Implicit Representations of Shapes with Boundaries

Multi-Scale Progressive Fusion Network for Single Image Deraining

The Curious Layperson: Fine-Grained Image Recognition without Expert Labels (BMVC 2021)