DETR

Detection Transformer (DETR) is an end-to-end object detection model implemented using the Transformer architecture.

Overview

Detection Transformer (DETR) is one of the first end-to-end object detection models implemented using the Transformer architecture. DETR was developed by Facebook Research. The baseline DETR model has an AP score of 42.0% on COCO.

It uses a transformer encoder-decoder architecture on top of a convolutional backbone (e.g., ResNet). This allows DETR to handle variable numbers of objects and complex object interactions more effectively than traditional CNN-based detectors.

By leveraging the power of transformers and attention mechanisms, DETR demonstrates improved generalization capabilities, handling diverse and complex scenes more effectively than traditional object detectors.

‍

Performance

	name	backbone	schedule	inf_time	box AP	url	size
0	DETR	R50	500	0.036	42.0	model \| logs	159Mb
1	DETR-DC5	R50	500	0.083	43.3	model \| logs	159Mb
2	DETR	R101	500	0.050	43.5	model \| logs	232Mb
3	DETR-DC5	R101	500	0.097	44.9	model \| logs	232Mb

Use This Model

Label Data Automatically with DETR

You can automatically label a dataset using DETR with help from Autodistill, an open source package for training computer vision models. You can label a folder of images automatically with only a few lines of code. Below, see our tutorials that demonstrate how to use DETR to train a computer vision model.

OWL-ViT and DETR

Grounding DINO and DETR

SAM-CLIP and DETR

Grounded SAM and DETR

DETIC and DETR