search

valeoai / LOST

Pytorch implementation of LOST unsupervised object discovery method
Other
236 stars 28 forks source link

Multi-object discovery #11

Closed rgbd-zml closed 2 years ago

rgbd-zml commented 2 years ago

HI, I have a confusion about the interesting work. How to perform multi-target discovery in the figure 1 (middle) of your paper？ Any advice is greatly appreciated.

osimeoni commented 2 years ago

Hi @rgbd-zml, Thank you for your interest. Our method LOST discovers one object per image. However, we subsequently train a detector in a fully-supervised fashion using the detections of LOST as pseudo ground-truth. Doing so we obtain a detector that outputs several predictions per image, which was used to obtain the middle figure in Figure 1 of our paper. I hope this helps!