Skip to content

Latest commit

 

History

History
73 lines (50 loc) · 2.96 KB

README_EN.MD

File metadata and controls

73 lines (50 loc) · 2.96 KB

FS2K-SDE Dataset

Chinese English

FS2K-SDE means FS2K sketch drawing episode。

This dataset is released based on @DengPingFan/FS2K, which mainly constructs a dataset from a small number of strokes to complete sketches for the face sketch images in it, i.e., face <--> sketch sequence.

This dataset can be well used for dynamic sketch face retrieval.

Download: Just download the zip package in the dataset folder in the repository directly.

Updates：

  • 2022/10/25： FS2K-SDE dataset released!

Introduction

We built a high quality face sketch retrieval dataset named FS2K-SDE.The dataset contains 2000+ face images and 14w+ sketch sequences, each face image corresponds to 70 face sketch sequences generated, as follows.

img

The dataset contains two subsets, dataset1 and dataset2, respectively.

Where dataset1 corresponds to photo1 and sketch1 in the Deng dataset; dataset2 corresponds to photo3 and sketch3 in the Deng dataset.

Specifically, in dataset1 and dataset2：

  • The contents of the sketch folder are the sketch sequence.
  • The contents of the cleaned folder are obtained by line simplification of the sketch provided with the Deng dataset.

ATTENTION: In accordance with the requirements of the open source license MIT, we do not provide original face images and original face sketches, for this part, please go to this repository DengPingFan/FS2K (github.com) to get them.

File Structure

FS2K-SDE
├── dataset1                    (1529, source: CASIA-WebFace)
│       ├── cleaned
│       ├── sketch              (10w+, corresponding to the FS2K sketch1)
|              |——train         (70% of sketch)
|              |——test          (30% of sketch)
├── dataset2
│       ├── cleaned             (477, source：free stock photos websites)
│       ├── sketch              (3w+, corresponding to the FS2K sketch2)
|              |——train         (70% of sketch)
|              |——test          (30% of sketch)
|——README.md

Experiment

SLFIR:

image-20221025130509813

Contact

This dataset is maintained by Dawei Dai (dw_dai@163.com) and his master's student Yutang Li (2018211556@stu.cqupt.edu.cn).

Citation

@INPROCEEDINGS{10095094, author={Dai, Dawei and Li, Yutang and Wang, Liang and Fu, Shiyu and Xia, Shuyin and Wang, Guoyin}, booktitle={ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, title={Sketch Less Face Image Retrieval: A New Challenge}, year={2023}, volume={}, number={}, pages={1-5}, keywords={Image retrieval;Signal processing;Acoustics;Task analysis;Speech processing;Faces;Sketch-Based Image Retrieval;Face Sketch;Partial Sketch;Sequence Learning}, doi={10.1109/ICASSP49357.2023.10095094}}