IMAGEPRODUCTS
Product and Packaging Image Sets
Product imagery across views, packaging surfaces and visual conditions for recognition, retrieval and packaging understanding.
View datasetIMAGEOBJECTS
Everyday Object Photo Sets
A set of object photographs becomes more useful when multiple views can be traced back to the same physical instance. Everyday Object Photo Sets preserve that identity relationship so viewpoint, state and appearance can be modelled without treating every photograph as a separate object.
The collection can also attach attributes or state information to the persistent object record, making it possible to study how the same instance changes across angle, condition or configuration. Detection or segmentation layers can be added when the task requires spatial supervision.
Technical structure
- Persistent instance identifiers across related images
- Viewpoint, state and attribute metadata where relevant
- Category information linked to the object record
- Optional bounding-box or segmentation annotations
Typical applications
Instance recognition, object retrieval, viewpoint robustness, state classification and visual representation learning.
Delivery note
Object categories, instance relationships and state vocabulary are defined according to the target visual task.
Delivery
The exact package depends on the collection scope. Where relevant, delivery can include task-specific records, manifests, provenance fields, stable identifiers, SHA-256 hashes, MLCommons Croissant 1.0 metadata and loading instructions.