We’ve released Image Search v2, which adds optional automatic image segmentation to Image Search. This lets shoppers upload a scene photo (for example, an outfit or a room) and get results for each detected item — instead of treating the whole image as a single query.

What’s enabled

• A new endpoint: POST >
• Optional segmentation mode (use_segmentation=true) that detects multiple objects in an image and searches each one
• Optional per-segment metadata in the response (return_segments=true) so clients can render selectable regions (bounding boxes / masks)
• Segment crop options: search_crop_type=bbox (default) or search_crop_type=segment (masked crop)

What it unlocks for you

• Multi-item discovery: Let shoppers go from “this whole photo” to “show me results for the jeans / shirt / shoes”.
• Better relevance on complex images: Reduced noise when the upload contains multiple products or background clutter.
• Stronger demo / evaluation flows: Make “shop from photo” experiences more interactive by letting users select what they mean within the image.

API docs

• Image Search v2 reference: https://docs.constructor.com/reference/v2-image-search-retrieve-search-results

If you have any questions or feedback, please connect with your Customer Success Manager.