├── README.md
├── config_CPC.py
├── config_GAICD.py
├── cropping_dataset.py
├── cropping_model.py
├── figures
├── motivation.jpg
├── pipeline.png
└── qualitative_comparison.png
├── generate_pseudo_heatmap.py
├── human_bboxes
├── CPC
│ ├── human_bboxes.json
│ └── image_crop.json
├── FCDB
│ ├── data_split.json
│ ├── human_bboxes.json
│ └── image_crop.json
├── FLMS
│ ├── human_bboxes.json
│ └── image_crop.json
└── GAICD
│ ├── human_bboxes.json
│ ├── human_data_split.json
│ ├── image_crop.json
│ └── original_data_split.json
├── requirements.txt
├── test.py
├── train_on_CPC.py
└── train_on_GAICD.py
/README.md:
--------------------------------------------------------------------------------
1 | # Human-Centric-Image-Cropping
2 |
3 | This is the official repository for the following paper:
4 |
5 | > **Human-centric Image Cropping with Partition-aware and Content-preserving Features** [[arxiv]](https://arxiv.org/pdf/2207.10269.pdf)
6 | >
7 | > Bo Zhang, Li Niu, Xing Zhao, Liqing Zhang
8 | > Accepted by **ECCV2022**.
9 |
10 | We consider a specific and practical application: human-centric image cropping, which focuses on the depiction of a person.
11 | To this end, we propose a human-centric image cropping method with human-centric partition and important content preservation.
12 | As illustrated in the figure below, the proposed method uses two novel feature designs for the candidate crop: partition-aware feature and content-preserving feature.
13 | The partition-aware feature allows to treat different regions in a candidate crop differently conditioned on the human information,
14 | and the content-preserving feature helps to preserve the important content to be included in a good crop.
15 |
16 |
17 |
18 |
25 |
26 |