Skip to content

Latest commit

 

History

History
25 lines (20 loc) · 3.07 KB

DATASET.md

File metadata and controls

25 lines (20 loc) · 3.07 KB
Raw

DATASET

Visual Genome

The following is adapted from Danfei Xu and neural-motifs.

Note that our codebase intends to support attribute-head too, so our VG-SGG.h5 and VG-SGG-dicts.json are different with their original versions in Danfei Xu and neural-motifs. We add attribute information and rename them to be VG-SGG-with-attri.h5 and VG-SGG-dicts-with-attri.json. The code we use to generate them is located at datasets/vg/generate_attribute_labels.py. Although, we encourage later researchers to explore the value of attribute features, in our paper "Unbiased Scene Graph Generation from Biased Training", we follow the conventional setting to turn off the attribute head in both detector pretraining part and relationship prediction part for fair comparison, so does the default setting of this codebase.

Download:

  1. Download the VG images part1 part2. Extract these images to the file datasets/vg/VG_100K. If you want to use other directory, please link it in DATASETS['VG_stanford_filtered']['img_dir'] of pysgg/config/paths_catelog.py.
  2. Download the scene graphs and extract them to datasets/vg/VG-SGG-with-attri.h5, or you can edit the path in DATASETS['VG_stanford_filtered_with_attribute']['roidb_file'] of pysgg/config/paths_catelog.py.
  3. Link the image into the project folder
ln -s /path-to-vg/VG_100K datasets/vg/stanford_spilt/VG_100k_images
ln -s /path-to-vg/VG-SGG-with-attri.h5 datasets/vg/stanford_spilt/VG-SGG-with-attri.h5

Openimage V4/V6

Download

The initial dataset(oidv6/v4-train/test/validation-annotations-vrd.csv) can be downloaded from offical website. The Openimage is a very large dataset, however, most of images doesn't have relationship annotations. To this end, we filter those non-relationship annotations and obtain the subset of dataset (.ipynb for processing ). You can download the processed dataset: Openimage V6(38GB), Openimage V4(28GB) The dataset dir contains the images and annotations folder. Link the open_image_v4 and open_image_v6 dir to the /datasets/openimages then you are ready to go.