OpenDataLake

Since it is all about data, this are data wrappers for common datasets which load the data into a common structure.

There are handlers for several datasets. To get you started quickly.

Install

Simply install it via pip.

pip install opendatalake

Classification

Here are all classification datasets. They are loaded as a generator spitting out a feature(-vector/image) and a one-hot-encoded label.

1. Named Folders (Foldername = Label)
2. MNIST
3. ImageNet [TODO]
4. Cifar10/Cifar100
5. LFW (named folders)
6. PASCAL VOC [TODO]
7. Places [TODO]

Segmentation

Here are all segmentation datasets. They are loaded as a generator spitting out a feature(-vector/image) and segmentation(-vector/image).

1. Coco (WIP)
2. CamVid [TODO]
3. Cityscapes [TODO]

Detection

Here are all detection datasets. They are loaded as a generator spitting out a feature(-vector/image) and a detections(-vector/image). A detection is either a detection_2d, detection_2.5d or detection_3d object defined here.

1. Bosch TLR
2. Coco (WIP)
3. Kitti Detection (incl. 3d)
4. Pascal Voc 3d (WIP)
5. Udacity Autti (WIP)
6. Udacity Crowdai (WIP)

Unlabeled

Here are all unlabeled datasets. They are loaded as a generator spitting out a feature(-vector/image).

1. Image Folder