search

DeepInsight-PCALab / Dataset-Catalog-

A catalog for `/data3`.
35 stars 9 forks source link

Datasets provided by Chuanwei #15

Closed deepcharle closed 6 years ago

deepcharle commented 6 years ago

Provided by ChuanweiZhou Dataset: FlyingChairs path:/data3/FlyingChairs/FlyingChairs size: 1.1T keywords: optical flow estimation homepage: https://lmb.informatik.uni-freiburg.de/resources/datasets/FlyingChairs.en.html description: including training and testing data along with those transformed into tensorflow tfrecords files

Dataset: FlyingThings path:/data3/FlyingThings size: 1.3T keywords: optical flow estimation homepage: https://lmb.informatik.uni-freiburg.de/resources/datasets/SceneFlowDatasets.en.html description: including training and testing data along with those transformed into tensorflow tfrecords files, where optical flow was only transformed from left view as well as in the order of into_future

dataset: MPI-Sintel path: /data3/MPI_Sintel size: 12G keywords: optical flow estimation homepage: http://sintel.is.tue.mpg.de/ description: including all training and testing data of MPI-Sintel dataset, especially for evaluating the performance of optical flow algorithms

Dataset: MOT path:/data3/MOT size: 1.9G keywords： multi object tracking homepage: https://motchallenge.net/ description: including training and testing data of mot dataset for multi object tracking, also including optical flow extracted by flownet2.0 if any help for tracking

dataset: vot path: /data3/VOT/vot2015 size: 1.4G keywords: single object tracking homepage: http://www.votchallenge.net/ description: including training and testing data of vot dataset for single object tracking, also including optical flow extracted by flownet2.0 if any help for tracking

dataset: Charades path: /data3/Charades size: 396G keywords: video recognition & video caption homepage: http://allenai.org/plato/charades/ description: including videos, flows of original size as well as those scaled into 480p, along with two-stream features computed by frames and optical flows in the frequency of 8fps. Evaluation codes were also provided

dataset: VID_detec path:/data3/Imagenet/ILSVRC2017/ size: 359G keywords: video detection homepage: http://www.image-net.org/challenges/LSVRC/ description: including all data from the Imagenet 2017 Challenge for video object detection, composed of DET and VID subsets.

dichen-cd commented 6 years ago

数据集的大小呢？

dichen-cd commented 6 years ago

FlyingChairs, FlyingThings, MPI-Sintel 三者关系混乱。如果是三个不同的数据集就请分开放置。并且后两者的数据集大小缺失。

out_flow 数据集缺少来源。

VID_detec 请归类到 /data3/ImageNet 目录下，并标明是17年的数据。如果跟ILSVRC2010, ILSVRC2012, ILSVRC2015 有重合的话请删除重复部分。

dichen-cd commented 6 years ago

OK. 写的很详细。