D

Detr Resnet 101 Panoptic

Developed by facebook
DETR is an end-to-end object detection model combining convolutional neural networks with Transformers, supporting panoptic segmentation tasks.
Downloads 610
Release Time : 3/2/2022

Model Overview

This model uses ResNet-101 as the backbone network, combined with Transformer architecture for end-to-end object detection and panoptic segmentation, trained on the COCO dataset.

Model Features

End-to-end training
No need for anchor box design in traditional object detection, directly outputs detection results
Transformer architecture
Utilizes self-attention mechanisms to process global contextual information
Bipartite matching loss
Uses Hungarian algorithm for optimal matching between predictions and annotations
Panoptic segmentation extension
Can be naturally extended to panoptic segmentation tasks by adding a mask head

Model Capabilities

Object detection
Instance segmentation
Panoptic segmentation
Image analysis

Use Cases

Computer vision
Scene understanding
Identify objects in images and their precise locations
Achieves 40.1 bounding box AP on COCO validation set
Autonomous driving
Object detection and segmentation in road scenes
Medical image analysis
Detection of organs or lesions in medical images
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase