Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

YouTube

DETR - End-to-End Object Detection with Transformers

Yannic Kilcher via YouTube

Overview

Explore an in-depth analysis of the DETR (DEtection TRansformer) paper, which introduces a revolutionary approach to object detection using transformers. Learn how this innovative method streamlines the detection pipeline by eliminating the need for hand-designed components like non-maximum suppression and anchor generation. Discover the key elements of the DETR framework, including its set-based global loss and transformer encoder-decoder architecture. Understand how the model uses learned object queries to reason about object relations and global image context, producing final predictions in parallel. Examine the architecture in detail, delving into the bipartite match loss function, object queries, and transformer properties. Compare DETR's performance to established baselines on the COCO dataset and explore its potential for panoptic segmentation. Gain insights into this conceptually simple yet powerful approach that achieves comparable accuracy and run-time performance to highly-optimized object detection models.

Syllabus

- Intro & High-Level Overview
- Problem Formulation
- Architecture Overview
- Bipartite Match Loss Function
- Architecture in Detail
- Object Queries
- Transformer Properties
- Results

Taught by

Yannic Kilcher

Reviews

Start your review of DETR - End-to-End Object Detection with Transformers

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.