Huy V. Vo
@huyvvo
Followers
62
Following
12
Statuses
19
Research Scientist at FAIR, Meta | Alumnus @inria_paris @valeoai @ENS_ULM @Ens_ParisSaclay (MVA) @Polytechnique
Paris
Joined October 2013
but its benefits are not limited to this single use case. It has potential applications in other problems where a uniform distribution of training / testing data points is a desirable property such as active learning, data pruning, improving sample / annotation efficiency, etc.
0
0
0
[5/5] It boosts the performance of weakly-supervised detectors significantly, narrows the performance gap to fully-supervised Fast RCNN, and demonstrate a better trade-off between performance and annotation efficiency than both weakly- and fully-supervised object detection.
0
0
1
[3/5] lag far behind fully-supervised counterparts. In this work, we attempt to reduce this gap by using active learning techniques to carefully select a few images, annotate them with bounding boxes, then fine-tune weakly-supervised object detectors with the new annotation.
0
0
0
I am proud to announce that our paper on Large-Scale Unsupervised Object Discovery is accepted at #NeurIPS2021. @evsizikova
Our new paper "Large-Scale Unsupervised Object Discovery" is on arxiv: We propose to use ranking methods for object discovery and show that our approach scales better than the baselines while yielding state-of-the-art results on COCO ...
0
1
4
RT @NYUDataScience: CDS MS students Tianshu Chu & Xinmeng Li, CDS Faculty Fellow Dr. Elena Sizikova @evsizikova, & colleagues have authored…
0
2
0