1. PoseCNN 贡献的YCB-Video Dataset ~ 265G
2. LineMod
3. SUN RGB-D
4. Pix3D
5. ScanNet
6. ObjectNet3D
7. PASCAL3D+
数据集开发语言:Matlab
包含标注工具 Annotation_tools
===========
There are 9 directories in this package:
- Anchor. Anchor points/landmarks of the 12 categories in .off format.
- Annotations. The augmented PASCAL and ImageNet annotations for the 12 categories.
- Annotation_tools. The annotation tools for bounding box and 3D annotation, and some scripts to visualize the annotations.
The scripts are helpful to understand the format of the annotation. - CAD. The CAD models used in the annotations both in .off and .mat formats.
- Images. The PASCAL and ImageNet images of the 12 categories.
- Image_sets. Separation of the ImageNet images into training and validation sets, and subtype names for each category.
- PASCAL. The original PASCAL VOC 2012 benchmark.
- Segmentation. Code for projecting CAD models using annotations to obtain the segmentation masks of objects.
- VDPM. Modified DPM [4] code for joint detection and pose estimation. The data directory contains our trained models and test results.
The entry functions for training and testing are pose_train.m and pose_test.m respectively.
The evaluation function is compute_recall_precision_accuracy.m.
@InProceedings{xiang_wacv14,
author = {Yu Xiang and Roozbeh Mottaghi and Silvio Savarese},
title = {Beyond PASCAL: A Benchmark for 3D Object Detection in the Wild},
booktitle = {IEEE Winter Conference on Applications of Computer Vision (WACV)},
year = {2014},
}