FE-Blurframe Dataset

✏️ Github   |   📑 Paper    |   🖼️ Viewer

This is the FE-Blurframe dataset, designed for motion-blurred image line segment detection with events.

Summary

The FE-Blurframe dataset is constructed using a dual-camera system with a DAVIS 346 event camera and a FLIR RGB camera mounted via a beam splitter. A total of 52 sequences were collected at Wuhan University across diverse indoor and outdoor scenes. Each sequence contains event streams and APS images from the event camera, along with synchronized high-frame-rate RGB images from the FLIR camera (200 Hz, 640×512), ensuring clear references for line segment annotation.

• File structure:

|-- events_raw
    |-- 0000001.npz
    |-- 0000002.npz
    |-- ...
|-- images-blur
    |-- 0000001.png
    |-- 0000002.png
    |-- ...
|-- images-start
    |-- 0000001.png
    |-- 0000002.png
    |-- ...
|-- images-end
    |-- 0000001.png
    |-- 0000002.png
    |-- ...
|-- train.jsonl
|-- test.jsonl

• Number of samples:
  • Train: 600
  • Test: 200

Download

• Download with huggingface-hub

python3 -m pip install huggingface-hub
huggingface-cli download --repo-type dataset lh9171338/FE-Blurframe --local-dir ./

• Download with Git

git lfs install
git clone https://huggingface.co/datasets/lh9171338/FE-Blurframe

Usage

• Load the dataset from Hugging Face Hub

from datasets import load_dataset

ds = load_dataset("lh9171338/FE-Blurframe", trust_remote_code=True)
print(ds)
# DatasetDict({
#     train: Dataset({
#         features: ['blur_image', 'start_image', 'end_image', 'events', 'H', 'image_size', 'junc', 'flow', 'lines', 'edges_positive'],
#         num_rows: 600
#     })
#     test: Dataset({
#         features: ['blur_image', 'start_image', 'end_image', 'events', 'H', 'image_size', 'junc', 'flow', 'lines', 'edges_positive'],
#         num_rows: 200
#     })
# })

• Load the dataset from local

from datasets import load_dataset

ds = load_dataset("FE-Blurframe", trust_remote_code=True)
print(ds)
# DatasetDict({
#     train: Dataset({
#         features: ['blur_image', 'start_image', 'end_image', 'events', 'H', 'image_size', 'junc', 'flow', 'lines', 'edges_positive'],
#         num_rows: 600
#     })
#     test: Dataset({
#         features: ['blur_image', 'start_image', 'end_image', 'events', 'H', 'image_size', 'junc', 'flow', 'lines', 'edges_positive'],
#         num_rows: 200
#     })
# })

• Load the dataset with jsonl files

import jsonlines

with jsonlines.open("test.jsonl") as reader:
    infos = list(reader)
print(infos[0].keys())
# dict_keys(['filename', 'image_size', 'H', 'junc', 'flow', 'lines', 'edges_positive'])

Citation

@ARTICLE{10323537,
  author={Yu, Huai and Li, Hao and Yang, Wen and Yu, Lei and Xia, Gui-Song},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
  title={Detecting Line Segments in Motion-Blurred Images With Events}, 
  year={2023},
  pages={1-16},
  doi={10.1109/TPAMI.2023.3334877}
}