D

Deformable Detr With Box Refine

Developed by SenseTime
Deformable DETR is an end-to-end object detection model that combines the advantages of Transformer architecture and deformable convolution, achieving efficient object detection on the COCO dataset.
Downloads 312
Release Time : 3/2/2022

Model Overview

This model adopts an encoder-decoder Transformer architecture, achieving efficient and accurate object detection through deformable attention mechanisms and bounding box refinement modules.

Model Features

Deformable attention mechanism
Effectively handles objects of different scales through deformable attention modules, improving detection accuracy
End-to-end training
Directly outputs detection results without complex post-processing steps
Bounding box refinement
Includes specialized bounding box refinement modules to enhance localization accuracy
Efficient Transformer architecture
Leverages the advantages of Transformers to achieve global context modeling

Model Capabilities

Object detection
Multi-category recognition
Bounding box prediction

Use Cases

Computer vision applications
Scene understanding
Identify and localize multiple objects in images
Accurately detects 80 categories in the COCO dataset
Intelligent surveillance
Real-time detection of targets in surveillance videos
Autonomous driving
Object detection in road scenes
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase