BLOG

BLOG

Computer vision: implementation and usage



Computer vision is found in nearly every aspect of our lives including images, videos, text and speech...
BLOG
Visual Search

The neural networks technology


The theory of neural networks creation started from the idea of designing intelligent computing devices like biological systems...

Data preparing for machine learning




There are three aspects for building a successful AI model: data, algorithms and calculations...
Visual Search

AI technologies

Machine learning, deep learning, computer vision and natural language processing are AI technologies, training computers to accomplish different kinds of tasks...

Our first client


How did we end up collecting photos at trash dumps?

The first image dataset we collected was for a customer 17,000 kilometers away.

Visual Search

The best source of training data is the real world

Training datasets are the knowledge and experience we feed into the model to make it smart.
Visual Search

Data Labeling Overview

Labeling is the first and the most important step in training and preparing data for machine learning.

Describe your task

Height Accuracy

Six steps for perfect bounding box annotation.
There is no standard template for a task description, so you just describe the task in your own words. Here are some examples.
Client Testimonials
Mehdi Mohseni
CEO and Founder at CellarEye, PhD
I am very happy with the quality of the annotations delivered by the Tagias team. Our object annotation task was pretty complicated. However, we were able to efficiently communicate the requirements with the project manager, and with very few iterations, the team was able to deliver annotations to meet and exceed our expectations. Highly recommended.
Alexey Grigorev
Lead Data Scientist at OLX Group
Tagias helped me collect an open image dataset of clothes. We collected over 5,000 images, 60% of which (3,000) were contributed by Tagias. The Tagias team took the pictures, and I didn't need to worry about copyright violations. The internal validation process was helpful: only a couple of images weren't suitable.