Researchers From Tel Aviv University, UC Berkeley and NVIDIA Introduce ‘DETReg’, A Novel Unsupervised AI For Object Detection

Date:

The application of self-supervised pretraining to computer vision has been beneficial, especially for object detection. However, previous approaches were not designed with localization in mind – which poses a key obstacle when it comes to tasks like detecting objects.

AI researchers have developed DETReg (DEtection with TRansformers based on Region priors), an innovative unsupervised pretraining approach for object detection. DETReg trains detectors with unlabeled data by providing two key pretraining tasks: The Object Localization Task and the Object Embedding Task. The first task aims to teach a model how to detect objects regardless of their type, while the latter is designed for understanding categories in images. With the simplicity of recent transformers for object detection, this research is chosen to base its approach on the Deformable DETR architecture, which simplifies implementation and training.

The researchers trained DETReg on the two tasks without manually annotated bounding boxes or categories. This approach allows all parameters to be learned, which will produce meaningful detections even with no supervision at all!

https://arxiv.org/pdf/2106.04550.pdf

The researchers ran their study with DETReg on standard benchmarks, MS COCO and PASCAL VOC. They found that it improved under all of the settings for these challenging baselines across the board. Particularly when there was less than a certain amount of annotated data available, they saw a performance improvement. DETReg outperforms previous supervised and unsupervised baseline approaches on low-data regime when trained with only 1%, 2%, 5%, and 10% of the labeled data on MS COCO.

https://arxiv.org/pdf/2106.04550.pdf

Paper: https://arxiv.org/pdf/2106.04550.pdf

Codes: https://github.com/amirbar/DETReg

Project: https://www.amirbar.net/detreg/

Asif Razzaqhttp://www.marktechpost.com
Asif Razzaq is an AI Journalist and Cofounder of Marktechpost, LLC. He is a visionary, entrepreneur and engineer who aspires to use the power of Artificial Intelligence for good. Asif's latest venture is the development of an Artificial Intelligence Media Platform (Marktechpost) that will revolutionize how people can find relevant news related to Artificial Intelligence, Data Science and Machine Learning. Asif was featured by Onalytica in it’s ‘Who’s Who in AI? (Influential Voices & Brands)’ as one of the 'Influential Journalists in AI' (https://onalytica.com/wp-content/uploads/2021/09/Whos-Who-In-AI.pdf). His interview was also featured by Onalytica (https://onalytica.com/blog/posts/interview-with-asif-razzaq/).

Share post:

Popular

More like this
Related

Recent Research on Manifolds in Commonly Used Atomic Fingerprints and Failure to Machine Learning Four-Body Interactions

Atomic fingerprints are often employed in machine learning situations...

A Neural Network for Solving and Generating University Level Mathematics Problems Using Program Synthesis

The AI research community widely believed that modern deep...

Increased Data Security Using ‘EzPC’ In The Machine Learning Model Validation Process

Artificial intelligence (AI) has revolutionized various industries in the...
Join the AI conversation and receive daily AI updates