Skip to content

Latest commit

 

History

History
23 lines (17 loc) · 1.96 KB

DATASET.md

File metadata and controls

23 lines (17 loc) · 1.96 KB
Raw

DATASET

The following is adapted from Scene-Graph-Benchmark, Danfei Xu and neural-motifs.

Download:

  1. Download the VG images part1 (9 Gb) part2 (5 Gb). 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 maskrcnn_benchmark/config/paths_catelog.py.
  2. Download the scene graphs labels 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 maskrcnn_benchmark/config/paths_catalog.py.
  3. Download the detection results of 3 datasets, including: Conceptual Caption, COCO Caption and Visual Genome. After downloading, you can run cat cc_detection_results.zip.part* > cc_detection_results.zip to merge several partitions into one zip file and unzip it to folder datasets/vg/.

Folder structure:

After downloading the above files, you should have following hierarchy in folder datasets/vg/:

├── VG_100K
├── cc_detection_results_oid
├── COCO_detection_results_oid
├── VG_detection_results_oid
└── VG-SGG-with-attri.h5

Preprocessing scripts

We provide scripts for data preprocessing, such as extracting the detection results from images and creating pseudo labels based on detection results and parsed concepts from image captions. More detail can be found in the folder preprocess.