Deta Resnet 50 24 Epochs
DETA is a transformer-based object detection model that significantly improves training efficiency and detection performance by reintroducing IoU assignment and non-maximum suppression techniques.
Downloads 27
Release Time : 1/30/2023
Model Overview
DETA is an innovative object detection model that combines transformer architecture with traditional computer vision techniques, achieving excellent detection accuracy while maintaining efficient training.
Model Features
Efficient Convergence
Achieves 50.2 mAP on the COCO dataset with only 12 training epochs.
Combination of Tradition and Innovation
Reintroduces IoU assignment and non-maximum suppression (NMS) techniques, leveraging the advantages of transformer architecture.
Training Efficiency
Training and testing speeds are comparable to Deformable DETR, but with faster convergence.
Model Capabilities
Object Detection
Image Analysis
Object Localization
Use Cases
Computer Vision
General Object Detection
Detects and locates multiple objects in various scenarios.
Achieves 50.2 mAP on the COCO dataset.
Real-time Surveillance Systems
Used for object detection and tracking in video surveillance.
Featured Recommended AI Models