CNN-Transformer-for-motion-prediciton
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Introduction This project implements a CNN + Transformer framework for image-sequence-based motion prediction of active micromotors.Given a sequence of images, the model predicts the motion state at each frame, including: cos(θ) sin(θ) normalized speed The project includes: model definition training pipeline custom sequence loss prediction and visualization training log export Model The model consists of three parts: CNN feature extractorResNet18 is used to extract spatial features from each frame. Transformer encoderTemporal dependencies across image sequences are modeled by a Transformer encoder. Regression headThe model outputs a 3-dimensional vector for each frame: [cos(theta), sin(theta), speed] Requirements This project requires the following environment and dependencies: Python 3.8+ PyTorch torchvision numpy pandas matplotlib opencv-python tqdm openpyxl
Fields of Study & Subjects
Micro and Nano Robotics
Advanced Neural Network Applications
Model Reduction and Neural Networks
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 27 March 2026
- Language
- EN
- Document type
- Dataset