In this premiere episode of Getty Images Tech Connect, our experts take on Visual Search, an important topic across the tech and media industries—filled with challenges and opportunities for developers. From embeddings to semantic understanding and diversity & inclusion, see how we’ve enabled our customers to find the images and videos they need, faster and easier, as well as explore themes they might not have previously considered—all through the power of process, evaluation, and model selection.

We’ll show you how we:

Help customers find similar images
Facilitate complex queries using embeddings
Eliminate bias in our modeling through continuous evaluation
Develop our algorithm for semantic understanding
Apply stylistic coverage for color, special effects, and apply histogram to drive the best search results

4:19 Intro
6:24 What is the customer need we are solving for?
8:14 How we use embedding to optimize visual search.
11:01 How we evaluate the models
19:38 How we brought similar images function to production
22:08 The results
24:45 Q&A
24:57 Q&A: How do you see search evolving to make all this easier for customers?
27:04 Q&A: Does scale of content make these problems even more challenging?
28:35 Q&A: What makes working at Getty Images in tech exciting?
31:01 Q&A: What are some of the main challenges you face in this work?
33:54 Q&A: Do some of the search models reinforce existing biases? How do we deal with that?
35:48 Q&A: How do you collaborate across the organization to achieve shared goals and serve the customer?

свернуть