Visual search accounts for 27 % of searches across ten major search properties.
Prominent brands such as eBay, The Home Depot, Alibaba Group, Forever 21, ASOS.com, Rakuten, India and many others have accelerated their visual‑search implementation.
The apparel‑detection service searches billions of images online to find visually similar fashion items.
- Detectable luxury brands include: Burberry, CHANEL, Fendi, Ralph Lauren, Prada Group.
I positioned the service under the name Labellerr, promoting it as a plug‑and‑play API that lets retailers quickly customize visual‑search workflows. See the broader retail‑focused annotation offering in retail focused data labeling platform.
📸 62 % of millennials prefer visual search to other modalities.
I proposed using a Multi‑View Convolutional Neural Network (MVCNN) together with part‑dimension metadata to improve classification and retrieval of 3‑D CAD models, aiming to cut the cost of manual tag annotation and enable better search and discovery of engineering models in large public repositories.
MVCNN algorithm + part dimension metadata resulting in better classification and retrieval of 3D computer‑aided design (CAD) models.
I also introduced a Context‑aware Visual Attention (CoVA) detection pipeline that mimics human ability to locate elements by combining appearance features from webpage screenshots with syntactical structure extracted from the DOM tree.
Mimic the ability of humans to detect the location of elements on a webpage with a Context‑aware Visual Attention‑based (CoVA) detection pipeline which combines a) appearance features b) syntactical structure from the DOM tree.
I highlighted a research paper that uses scene graphs and a conditional random field model to enable detailed semantic image retrieval, introducing a dataset of 5 000 human‑annotated scene graphs.