DATASET EXPLORATION

Creating a new dataset

5min
inside dataset exploration, datasets, and their items can be explored using our powerful gallery docid\ rpya vhea8snh5bi6gaev and metrics docid 1 sdbefhejsskc2qb0bwc views a new dataset can be created from the dataset exploration page defining the dataset content and how its displayed when defined, the dataset will contain a specific type of annotation created for a specific set of scenes these annotated scenes will be displayed as items in dataset exploration specifying scenes you decide which annotated scenes (specific images, sequences, or point clouds) should be part of the dataset by first selecting the annotation projects they are part of and later selecting which taxonomy their annotations were created with the included scenes can be further limited by selecting to exclude annotated scenes from specific annotation requests specifying the type of annotations the type of annotations that are included in the dataset is defined by their annotation taxonomy the annotation taxonomy (in some instances equal to annotation instruction) is what defines what type of geometries and attributes that are annotated in the scenes only taxonomies used in all defined projects can be selected deciding how to display the dataset all datasets are represented as items inside dataset exploration these items are represented as individual thumbnails in the gallery docid\ rpya vhea8snh5bi6gaev when creating a new dataset you define which type of item the dataset should be displayed as by selecting a dataset type object centric datasets for object detection datasets, items are object centric , meaning that each item represents an individual object in one frame for sequence data, each item consists of all frames where a specific object is labeled if your scenes contain single frame data, you can select between the types single frame 2d box and single frame 3d box if your scenes contain sequence data, you can only select the type sequence 3d box frame centric datasets for frame centric datasets, where attributes are applicable for the full frame, you can select the type single frame frame properties for these datasets items represent one frame segmentation datasets in segmentation datasets, items represent a frame these datasets can currently only be created manually by kognic reach out if you need help creating such a dataset creating the dataset after specifying the content of the dataset and how it should be displayed, you can create it when created, kognic starts the process of creating the dataset items depending on the size of the dataset the time it takes will vary if you think it is taking a suspicious amount of time, reach out to kognic, and we will make sure things are working as expected