An implementation of Equivariant e2 convolutional kernals into a convolutional self attention network, applied to radio astronomy data.

Last update: Nov 09, 2021

Related tags

Deep Learning EquivariantSelfAttention

Overview

EquivariantSelfAttention

An implementation of Equivariant e2 convolutional kernals into a convolutional self attention network, applied to radio astronomy data.

This work extends and builds on previous work:

FR Classification using e2cnn, whose code is here.
Attention for FR Classificaion, whose code is here. This work makes extensive use of the e2cnn package: e2cnn.

Imports for General Use

import os
import torch
import e2cnn
import pandas as pd
import numpy as np
import configparser as ConfigParser

import utils

Loading a network:

To load a network import the network files, read in a config file, and load the network with the config file parameters:

from networks import AGRadGalNet, VanillaLeNet, testNet, DNSteerableLeNet, DNSteerableAGRadGalNet
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

config      = ConfigParser.ConfigParser(allow_no_value=True)
config_path = "configs/CONFIG_NAME"
config.read(config_path)

net = locals()[config['model']['base']](**config['model']).to(device)

Alternatively, if you know which architecture you want to use, only import that and load in the config parameters (i.e. without using locals()):

from networks import DNSteerableAGRadGalNet
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

config      = ConfigParser.ConfigParser(allow_no_value=True)
config_path = "configs/CONFIG_NAME"
config.read(config_path)

net = DNSteerableAGRadGalNet(**config['model']).to(device)

Loading a model:

To load the best performing model (according to early stopping) from a given model training session:

path_supliment = config['data']['augment']+'/'
model = utils.utils.load_model(config, load_model='best', device=device, path_supliment=path_supliment)

Testing a model:

Follow the example of evaluation.ipynb for model evaluation.

Defining a model:

To define your own model simply create a new config which follows the examples provided in configs.

Training a model:

To train a model use the train.py using your config file:

python3.8 train.py --config YOUR_CONFIG_NAME.cfg >& logs/YOUR_CONFIG_NAME.log

An implementation of Equivariant e2 convolutional kernals into a convolutional self attention network, applied to radio astronomy data.

Related tags

Overview

EquivariantSelfAttention

Imports for General Use

Loading a network:

Loading a model:

Testing a model:

Defining a model:

Training a model:

Owner

LETR: Line Segment Detection Using Transformers without Edges

This is the first released system towards complex meters` detection and recognition, which is implemented by computer vision techniques.

Unsupervised Foreground Extraction via Deep Region Competition

Medical Insurance Cost Prediction using Machine earning

SHRIMP: Sparser Random Feature Models via Iterative Magnitude Pruning

Awesome Human Pose Estimation

MODALS: Modality-agnostic Automated Data Augmentation in the Latent Space

Rotation Robust Descriptors

Capstone-Project-2 - A game program written in the Python language

Contains code for Deep Kernelized Dense Geometric Matching

Expressive Power of Invariant and Equivaraint Graph Neural Networks (ICLR 2021)

A Shading-Guided Generative Implicit Model for Shape-Accurate 3D-Aware Image Synthesis

[ECCV'20] Convolutional Occupancy Networks

Official Pytorch implementation of "CLIPstyler:Image Style Transfer with a Single Text Condition"

nnFormer: Interleaved Transformer for Volumetric Segmentation Code for paper "nnFormer: Interleaved Transformer for Volumetric Segmentation "

Cross-media Structured Common Space for Multimedia Event Extraction (ACL2020)

Code for the paper "PortraitNet: Real-time portrait segmentation network for mobile device" @ CAD&Graphics2019

[CVPR2022] Bridge-Prompt: Towards Ordinal Action Understanding in Instructional Videos

Object Detection Projekt in GKI WS2021/22

Rot-Pro: Modeling Transitivity by Projection in Knowledge Graph Embedding