CNN-Transformer-for-motion-prediciton

Fuente: Zenodo
Enregistré dans:
Détails bibliographiques
Auteur principal: Liu, Jiaxin
Format: Recurso digital
Langue:anglais
Publié: Zenodo 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866901884293349376
author Liu, Jiaxin
author_facet Liu, Jiaxin
contents <div dir="auto"> <h2>Introduction</h2> <a href="https://github.com/husandaimei/CNN-Transformer-for-motion-prediciton#introduction"></a></div> <p dir="auto">This project implements a <strong>CNN + Transformer</strong> framework for image-sequence-based motion prediction of active micromotors.<br>Given a sequence of images, the model predicts the motion state at each frame, including:</p> <ul> <li><code>cos(θ)</code></li> <li><code>sin(θ)</code></li> <li><code>normalized speed</code></li> </ul> <p dir="auto">The project includes:</p> <ul> <li>model definition</li> <li>training pipeline</li> <li>custom sequence loss</li> <li>prediction and visualization</li> <li>training log export</li> </ul> <div dir="auto"> <h2>Model</h2> <a href="https://github.com/husandaimei/CNN-Transformer-for-motion-prediciton#model"></a></div> <p dir="auto">The model consists of three parts:</p> <ol> <li> <p dir="auto"><strong>CNN feature extractor</strong><br>ResNet18 is used to extract spatial features from each frame.</p> </li> <li> <p dir="auto"><strong>Transformer encoder</strong><br>Temporal dependencies across image sequences are modeled by a Transformer encoder.</p> </li> <li> <p dir="auto"><strong>Regression head</strong><br>The model outputs a 3-dimensional vector for each frame:</p> <div> <pre><code>[cos(theta), sin(theta), speed]</code></pre> </div> </li> </ol> <div dir="auto"> <h2>Requirements</h2> <a href="https://github.com/husandaimei/CNN-Transformer-for-motion-prediciton#requirements"></a></div> <p dir="auto">This project requires the following environment and dependencies:</p> <ul> <li>Python 3.8+</li> <li>PyTorch</li> <li>torchvision</li> <li>numpy</li> <li>pandas</li> <li>matplotlib</li> <li>opencv-python</li> <li>tqdm</li> <li>openpyxl</li> </ul>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19245390
institution Zenodo
language eng
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle CNN-Transformer-for-motion-prediciton
Liu, Jiaxin
microrobot
micromanipulation
Janus micromotor
<div dir="auto"> <h2>Introduction</h2> <a href="https://github.com/husandaimei/CNN-Transformer-for-motion-prediciton#introduction"></a></div> <p dir="auto">This project implements a <strong>CNN + Transformer</strong> framework for image-sequence-based motion prediction of active micromotors.<br>Given a sequence of images, the model predicts the motion state at each frame, including:</p> <ul> <li><code>cos(θ)</code></li> <li><code>sin(θ)</code></li> <li><code>normalized speed</code></li> </ul> <p dir="auto">The project includes:</p> <ul> <li>model definition</li> <li>training pipeline</li> <li>custom sequence loss</li> <li>prediction and visualization</li> <li>training log export</li> </ul> <div dir="auto"> <h2>Model</h2> <a href="https://github.com/husandaimei/CNN-Transformer-for-motion-prediciton#model"></a></div> <p dir="auto">The model consists of three parts:</p> <ol> <li> <p dir="auto"><strong>CNN feature extractor</strong><br>ResNet18 is used to extract spatial features from each frame.</p> </li> <li> <p dir="auto"><strong>Transformer encoder</strong><br>Temporal dependencies across image sequences are modeled by a Transformer encoder.</p> </li> <li> <p dir="auto"><strong>Regression head</strong><br>The model outputs a 3-dimensional vector for each frame:</p> <div> <pre><code>[cos(theta), sin(theta), speed]</code></pre> </div> </li> </ol> <div dir="auto"> <h2>Requirements</h2> <a href="https://github.com/husandaimei/CNN-Transformer-for-motion-prediciton#requirements"></a></div> <p dir="auto">This project requires the following environment and dependencies:</p> <ul> <li>Python 3.8+</li> <li>PyTorch</li> <li>torchvision</li> <li>numpy</li> <li>pandas</li> <li>matplotlib</li> <li>opencv-python</li> <li>tqdm</li> <li>openpyxl</li> </ul>
title CNN-Transformer-for-motion-prediciton
topic microrobot
micromanipulation
Janus micromotor
url https://doi.org/10.5281/zenodo.19245390