Top 2020 Computer Vision Papers Highlight Transformers and 3D Innovation
Researchers explore whether transformers will revolutionize computer vision just as they transformed natural language processing. A new curated list highlights the top ten 2020 papers covering object detection, 3D modeling, and generative networks.
Computer vision researchers in 2020 actively explore whether transformers will revolutionize image recognition just as they transformed natural language processing. Early results show that transformers achieve highly promising results on vision tasks, sparking significant interest across the artificial intelligence community. Alongside this transformer focus, scientists continue to make major strides in learning 3D objects from images and generating realistic visuals using GANs and autoencoders.
To help navigate this overwhelming volume of research, experts curate and summarize the top ten computer vision papers from the year. This essential list serves as a strong starting point for anyone wanting to understand the latest trends and breakthroughs in the field. The featured research covers a wide array of complex topics, including scalable object detection, optical flow, and adversarial latent autoencoders.
One standout paper on this list introduces EfficientDet, a new family of object detectors that prioritizes model efficiency without sacrificing accuracy. EfficientDet utilizes a weighted bi-directional feature pyramid network for fast multi-scale feature fusion and applies a compound scaling method to uniformly scale resolution, depth, and width. Built upon EfficientNet backbones, these optimized detectors consistently achieve state-of-the-art performance.