English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multi-scale deep CNN architecture [1, 2, 3] successfully captures both fineand coarse level image descriptors for visual similarity task, but they come upwith expensive memory overhead and latency. In this paper, we propose acompeting novel CNN architecture, called MILDNet, which merits by being vastlycompact (about 3 times). Inspired by the fact that successive CNN layersrepresent the image with increasing levels of abstraction, we compressed ourdeep ranking model to a single CNN by coupling activations from multipleintermediate layers along with the last layer. Trained on the famousStreet2shop dataset [4], we demonstrate that our approach performs as good asthe current state-of-the-art models with only one third of the parameters,model size, training time and significant reduction in inference time. Thesignificance of intermediate layers on image retrieval task has also been shownto be performing on popular datasets Holidays, Oxford, Paris [5]. So eventhough our experiments are done on ecommerce domain, it is applicable to otherdomains as well. We further did an ablation study to validate our hypothesis bychecking the impact on adding each intermediate layer. With this we alsopresent two more useful variants of MILDNet, a mobile model (12 times smaller)for on-edge devices and a compactly featured model (512-d feature embeddings)for systems with less RAMs and to reduce the ranking cost. Further we presentan intuitive way to automatically create a tailored in-house triplet trainingdataset, which is very hard to create manually. This solution too can also bedeployed as an all-inclusive visual similarity solution. Finally, we presentour entire production level architecture which currently powers visualsimilarity at Fynd.

MILDNet: A Lightweight Single Scaled Deep Ranking Architecture