On 2024-07-11 17:30 UTC I recorded an interesting example of CLIP's limitations illustrated in the paper “Active Open‑Vocabulary Recognition: Let Intelligent Moving Mitigate CLIP Limitations” by Jianxiong Zhou, Xiaoying Xing, Ying Wu, and Lei Fan.
I am investigating the PROB model, which was published in CVPR 2023, for open‑set object detection to accelerate labeling pipelines.
PROB was actually published in CVPR 2023.
My Labellerr team – Simran bajaj, Ximi H., and I – are experimenting with PROB for open‑set detection.
The goal is to detect unknown objects automatically to speed up human labeling pipelines.