Models

Model catalog for the current HARENA suite. Params and MACs are means across the datasets for 3 s windows.

Snapshot
Models
19

Release Timeline

16/16 dated models

2015 2017 2019 2021 2023 2025 2026

Filters

Building Blocks
All building blocks 7/7
Framework
All frameworks 2/2
Params
0 - 3M range
Min 0 Max 3M
MACs
0 - 610M range
Min 0 Max 610M

19 Models

Name Building Blocks
LITEWAY Affiliations withheld during review 2026
Attention ⓘ Convolutional ⓘ
PyTorch 5.6k 724.9k
LITEWAY-Light Affiliations withheld during review 2026
Attention ⓘ Convolutional ⓘ
PyTorch 5.4k 615.3k
TinierHAR Affiliations withheld during review 2025
Attention ⓘ Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 22.0k 1.57M
MLP-HAR Affiliations withheld during review 2024
Dense ⓘ Spectral ⓘ
PyTorch 1.43M 27.37M
MLP-Mixer Affiliations withheld during review 2023
Dense ⓘ
PyTorch 830.9k 188.88M
DynamicWHAR Affiliations withheld during review 2022
Convolutional ⓘ Dense ⓘ Graph ⓘ
PyTorch 2.98M 305.81M
TinyHAR Affiliations withheld during review 2022
Attention ⓘ Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 252.8k 25.64M
Triple-Cross-Attn Affiliations withheld during review 2022
Attention ⓘ Convolutional ⓘ Dense ⓘ
PyTorch 766.7k 50.77M
Attend-and-Discriminate Affiliations withheld during review 2021
Attention ⓘ Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 836.0k 442.74M
DANA Affiliations withheld during review 2021
Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 1.08M 273.82M
DeepConvShallowLSTM Affiliations withheld during review 2021
Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 1.02M 601.40M
SA-HAR Affiliations withheld during review 2020
Attention ⓘ Convolutional ⓘ Dense ⓘ
PyTorch 424.7k 107.93M
DeepConvLSTM-Attn Affiliations withheld during review 2018
Attention ⓘ Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 1.09M 410.27M
Guan-LSTM Affiliations withheld during review 2017
Dense ⓘ Recurrent ⓘ
PyTorch 819.1k 146.56M
DeepConvLSTM Affiliations withheld during review 2016
Convolutional ⓘ Dense ⓘ Recurrent ⓘ
PyTorch 1.08M 407.58M
CNN-HAR Affiliations withheld during review 2015
Convolutional ⓘ
PyTorch 216.9k 30.27M
k-NN Affiliations withheld during review · k=5, distance-weighted -
Handcrafted ⓘ
scikit-learn - -
Random Forest Affiliations withheld during review · 100 trees, depth=8 -
Handcrafted ⓘ
scikit-learn - -
SVM Affiliations withheld during review · RBF kernel, C=1 -
Handcrafted ⓘ
scikit-learn - -