VideoContentSearchMultimodalTopicTrainingFeatures

GoogleApi.ContentWarehouse.V1.Model.VideoContentSearchMultimodalTopicTrainingFeatures

Table of Contents ▼

Jump to a specific part of the page:

Multimodal features for a single generated topic used to build training data.

This module has the following attributes (case-insensitive ascending order):

View Attributes

maxFrameSimilarityInterval (type: GoogleApi.ContentWarehouse.V1.Model.VideoContentSearchFrameSimilarityInterval, default: nil)
- The similarity info for the frame with maximum similarity to the topic in its visual interval. The repeated similarity field in this proto has a single value corresponding to the maximum similarity. This similarity score is used to filter and pick the training data examples.
normalizedTopic (type: String.t, default: nil)
- The topic/query normalized for Navboost and QBST lookups as well as fetching of the Rankembed nearest neighbors.
qbstTermsOverlapFeatures (type: GoogleApi.ContentWarehouse.V1.Model.VideoContentSearchQbstTermsOverlapFeatures, default: nil)
- QBST terms overlap features for a candidate query.
rankembedNearestNeighborsFeatures (type: GoogleApi.ContentWarehouse.V1.Model.VideoContentSearchRankEmbedNearestNeighborsFeatures, default: nil)
- Rankembed similarity features for a candidate nearest neighbor rankembed query.
saftEntityInfo (type: GoogleApi.ContentWarehouse.V1.Model.VideoContentSearchSaftEntityInfo, default: nil)
- The information about the saft entity annotation for this topic.
topicDenseVector (type: list(number()), default: nil)
- Raw float feature vector of the topic's co-text embedding representation in the Starburst space.

@spec decode(struct(), keyword()) :: struct()