University of Washington Liquid Pouring Dataset (UWLPD)
The University of Washington Liquid Pouring Dataset (UWLPD) contains 648 experiments. Each experiment consists of a sequence of egocentric RGB images captured at 30 FPS of a robot pouring liquid from one container to another. For each RGB image, there is a corresponding ground truth image containing the pixel-wise labels for the location of the liquid in the image in the blue channel.
The dataset is split into 18 zip files listed below, each approzimately 9-16 GB in size. The entire dataset totals 208 GB. Click on the links below to download each of the 18 files.
- scene_left_bowl_bottle.zip[16.2GB]
- scene_left_bowl_cup.zip[12.5GB]
- scene_left_bowl_mug.zip[10.0GB]
- scene_left_fruitBowl_bottle.zip[10.3GB]
- scene_left_fruitBowl_cup.zip[14.1GB]
- scene_left_fruitBowl_mug.zip[9.3GB]
- scene_left_pan_bottle.zip[10.0GB]
- scene_left_pan_cup.zip[12.1GB]
- scene_left_pan_mug.zip[9.3GB]
- scene_right_bowl_bottle.zip[15.9GB]
- scene_right_bowl_cup.zip[11.6GB]
- scene_right_bowl_mug.zip[11.2GB]
- scene_right_fruitBowl_bottle.zip[13.4GB]
- scene_right_fruitBowl_cup.zip[13.2GB]
- scene_right_fruitBowl_mug.zip[10.8GB]
- scene_right_pan_bottle.zip[14.2GB]
- scene_right_pan_cup.zip[13.3GB]
- scene_right_pan_mug.zip[10.6GB]
If you use this dataset in a publication, please cite the following:
Schenck, C., Fox, D., "Perceiving and Reasoning About Liquids Using Fully Convolutional Networks," In International Journal of Robotics Research (IJRR), Vol 37, No 4-5, pp. 452-471, April, 2018.