Active Event Alignment for Monocular Distance Estimation

This is the implementation of the paper "Active Event Alignment for Monocular Distance Estimation". The code takes events as input.

link to the paper: https://openaccess.thecvf.com/content/WACV2025/html/Cai_Active_Event_Alignment_for_Monocular_Distance_Estimation_WACV_2025_paper.html

Authors: Nan Cai, Pia Bideau

Setup

High-level Input-Output

Input:

• Events.
• Semantic segmentation mask
• IMU angular velocity

Output:

• Relative distance of different objects (region)

Usage

Requirements

pandas==1.0.5
matplotlib==3.2.1
numpy==1.18.5
opencv_python==4.2.0.34
torch==1.8.0
scikit_learn==0.24.2
scipy==1.4.1

Download datasets

• EVIMO2 dataset: In our paper we use samsung_mono sequences of the Structure from Motion type.
  • we include an example EVIMO2 data for the test

Test

First, run the ST-PPP algorithm to estimate the "virtual" angular velocity of the event camera for every object in the segmentation mask:

python3 ./src/main.py --dataset evimo \
--path ./dataset \
--save ./stppp_output \
--load_mask True \
--seq 4 --apply_mask True -f --dtime 0.05 --iter 100

Second, loop through all the object IDs using the following script:

sh obj.sh

Then, use the ST-PPP results to estimate the relative distance of each object:

python3 ./relative_distance/relative_dis_kalman.py \
--seq 4 \
--path ./dataset \
--estimate ./stppp_output

The relative depth figure would be saved to save_dir/current_date/sequence

Result Structure

rel_res_frame_dict.npy

This file contains a dictionary where each key is a timestamp (formatted as a string) and each value is another dictionary. The inner dictionary maps object IDs to their relative distances at that specific timestamp.

Example structure:

{
    "0.000000": {
        5000: 1.0,
        6000: 0.8,
        ...
    },
    "0.050000": {
        5000: 1.1,
        6000: 0.85,
        ...
    },
    ...
}

rel_res_obj_dict.npy

This file contains a dictionary where each key is an object ID and each value is another dictionary. The inner dictionary maps timestamps (formatted as strings) to the relative distances of the object at those specific timestamps.

Example structure:

{
    5000: {
        "0.000000": 1.0,
        "0.050000": 1.1,
        ...
    },
    6000: {
        "0.000000": 0.8,
        "0.050000": 0.85,
        ...
    },
    ...
}

Acknowledgements

This code leverages the following repository for impelementation of ST-PPP algorithm:

• ST-PPP