search

WongKinYiu / yolor

implementation of paper - You Only Learn One Representation: Unified Network for Multiple Tasks (https://arxiv.org/abs/2105.04206)
GNU General Public License v3.0
1.98k stars 524 forks source link

Custom Training with negative samples, possible with YOLOR? #257

Open SanKumSan opened 2 years ago

SanKumSan commented 2 years ago

Hello , I would like to train YoloR for custom object detection, with negative samples (Images with no objects and the corresponding text file will be empty).

  1. Is this approach followed in YoloR or is it a good idea to do so, because, the other YOLO versions supports this learns well and decreases the FP from these negative samples.
  2. And , if its possible, what all I have to do, keeping the data already ready with positive and negative samples. ?
  3. The code throws the assertion error, when the input image has no labels text file. Any changes need to be done ?

Regards

ramdhan1989 commented 1 year ago

Hi, I wonder which yolo that can use negative samples? if possible, would you mind telling me the repository? I am struggle to reduce FP in my case.