Datasets:

Juanxi
/

Phys101

@@ -1,129 +0,0 @@
-===============================================================================
-The Physics 101 Dataset (release v1.0, Jan 4, 2017)
-===============================================================================
-http://phys101.csail.mit.edu
-We introduce a new type of dataset – one that captures physical interactions
-of objects. The dataset consists of five different scenarios of 101 objects
-made of different materials and with a variety of masses and volumes.
-===============================================================================
-Dataset: Objects
-===============================================================================
-The dataset involves 101 objects of 15 materials/categories:
-- cardboard (cardboard)
-- dough (dough)
-- foam (foam)
-- hollow rubber (h_rubber)
-- hollow wood (h_wood)
-- metal coin (m_coin)
-- metal pole (m_pole)
-- plastic block (p_block)
-- plastic doll (p_doll)
-- plastic ring (p_ring)
-- plastic toy (p_toy)
-- porcelain (porcelain)
-- rubber (rubber)
-- wooden block (w_block)
-- wooden pole (w_pole)
-'objects.html' is a webpage showing all objects.
-In 'objects', file 'mass' lists the mass of objects (in grams), and 'vol' lists
-the volume of objects (in milliliters).
-===============================================================================
-Dataset: Scenarios
-===============================================================================
-The dataset consists of five scenarios.
-1) Ramp: We put an object on an inclined surface, and the object may either
-slide down or keep static, due to gravity and friction.
-The ramp scenario has four different settings:
-a) 10_01: The angle between the ramp and the table is 10 degree, and there is
-a green cardbox on the table.
-b) 10_02: The angle between the ramp and the table is 10 degree, and there is
-a piece of styrofoam on the table.
-c) 20_01: The angle between the ramp and the table is 20 degree, and there is
-a green cardbox on the table.
-d) 20_02: The angle between the ramp and the table is 20 degree, and there is
-a piece of styrofoam on the table.
-2) Spring: We hang objects on a spring, and gravity on the object will stretch
-the spring.
-The spring scenario has two different settings:
-a) loose: We hang objects on a looser spring.
-b) tight: We hang objects on a tighter spring.
-3) Fall: We drop objects in the air, and they freely fall onto various
-surfaces.
-The fall scenario has five different settings:
-a) foam: We drop objects on a piece of foam.
-b) glass: We drop objects on a piece of glass.
-c) metal: We drop objects on a metal board.
-d) rug: We drop objects on a rug.
-e) table: We drop objects on the table directly.
-4) Liquid: We drop objects into water, and they may float or sink at various
-speeds. The liquid scenario has a single setting.
-5) Multi: Similar to the ramp scenario, but with three objects in a row.
-The multi scenario has two different settings:
-a) 01_02: The first object on the table is a green cardbox, and the second is
-a piece of styrofoam.
-b) 02_01: The first object on the table is a piece of styrofoam, and the
-second is a green cardbox.
-===============================================================================
-Dataset: Videos
-===============================================================================
-The dataset consists of multiple trials for each setting, and each trial
-consists of the following temporally aligned videos:
-- Camera_1.mp4: front view RGB video taken by a DSLR camera
-- Camera_2.mp4: side view RGB video taken by a DSLR camera
-- Kinect_FullDepth_1.mp4: upper front view depth video taken by a Kinect V2
-- Kinect_RGB_1.mp4: upper front view RGB video taken by a Kinect V2
-- Kinect_RGB-D_1.mp4: Kinect_RGB_1.mp4 registered with Kinect_Depth_1.mp4
-This release (v1.0) includes 3,038 trials in total. There are thus 3,038 x 5 =
-15,190 videos.
-===============================================================================
-Publications
-===============================================================================
-If you use our dataset in a publication, please kindly cite
-@inproceedings{wu2016physics,
-  title={Physics 101: Learning Physical Object Properties from Unlabeled Videos},
-  author={Wu, Jiajun and Lim, Joseph J and Zhang, Hongyi and Tenenbaum, Joshua B and Freeman, William T},
-  booktitle={BMVC},
-  year={2016}
-}
-The following closely related paper also used part of the Physics 101 dataset:
-@inproceedings{wu2015galileo,
-  title={Galileo: Perceiving physical object properties by integrating a physics engine with deep learning},
-  author={Wu, Jiajun and Yildirim, Ilker and Lim, Joseph J and Freeman, William T and Tenenbaum, Joshua B},
-  booktitle={NIPS},
-  year={2015}
-}
-===============================================================================
-Contact: Jiajun Wu, jiajunwu@mit.edu
-===============================================================================