ELP-One-step-Event-driven-AF

This repository contains the code and resources for One-Step Event-Driven High-Speed Autofocus (CVPR2025), which evaluates the ELP method on three datasets.

Dataset and Project Download

• Download the datasets from the following link:
  Datasets on Google Drive

• Clone this project repository:

  git clone https://github.com/YuHanBaozju/ELP.git
  cd ELP-One-step-Event-driven-AF
  

Testing ELP on Three Datasets

1. DAVIS Dataset

Navigate to the DAVIS_ELP directory and run:

cd DAVIS_ELP
python eval_davis.py

Note:
Before running, make sure to modify the dataset root directory path inside eval_davis.py to point to your downloaded DAVIS dataset.

2. EVK4 Dataset

Navigate to the EVK4_ELP directory and run:

cd EVK4_ELP
python eval_evk4.py

Note:
Modify the dataset root directory path inside eval_evk4.py accordingly.

3. Synthetic Dataset

Navigate to the synthetic_ELP directory and run:

cd synthetic_ELP
python eval_sim.py

Note:
Modify the dataset root directory path inside eval_sim.py accordingly.

Visualization of Results

To visualize the ELP method’s performance on different datasets:

• Set vis=True in each of the corresponding eval_xxx.py files:

Requirements

opencv-python numpy matplotlib

Citation

#TBD