- توضیحات :
مجموعه داده های NYU-Depth V2 شامل توالی های ویدئویی از صحنه های مختلف داخلی است که توسط هر دو دوربین RGB و Depth از Microsoft Kinect ضبط شده است.
صفحه اصلی : https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html
کد منبع :
tfds.image.NyuDepthV2
نسخه ها :
-
0.0.1
(پیش فرض): بدون یادداشت انتشار.
-
اندازه بارگیری :
31.92 GiB
اندازه مجموعه داده :
74.03 GiB
ذخیره خودکار ( مستندات ): خیر
تقسیم :
شکاف | مثال ها |
---|---|
'train' | 47،584 |
'validation' | 654 |
- ویژگی ها :
FeaturesDict({
'depth': Tensor(shape=(480, 640), dtype=tf.float16),
'image': Image(shape=(480, 640, 3), dtype=tf.uint8),
})
کلیدهای تحت نظارت (به
as_supervised
سند نظارت شده مراجعه کنید):('image', 'depth')
نقل قول :
@inproceedings{Silberman:ECCV12,
author = {Nathan Silberman, Derek Hoiem, Pushmeet Kohli and Rob Fergus},
title = {Indoor Segmentation and Support Inference from RGBD Images},
booktitle = {ECCV},
year = {2012}
}
@inproceedings{icra_2019_fastdepth,
author = {Wofk, Diana and Ma, Fangchang and Yang, Tien-Ju and Karaman, Sertac and Sze, Vivienne},
title = {FastDepth: Fast Monocular Depth Estimation on Embedded Systems},
booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
year = {2019}
}
- شکل ( tfds.show_examples ):
- مثالها ( tfds.as_dataframe ):