Skip to main content

New answers tagged

0 votes

Training object detection models from the scratch without using pre-trained models and weights

It is possible logically, but the accuracy might be low. As you mentioned, the purpose of this work is to study architecture. I agree with @Pranshu Mishra that image preprocessing, like augmentation, ...
Rosseta's user avatar
0 votes

Why YOLO algorithm predicts B boxes for each grid cell S?

Usually it is done in the context of Anchors. The idea is to enable detecting more than 1 object within the same cell: How would you detect both the person and the car? By adding an assumption that ...
Royi's user avatar
  • 195

Top 50 recent answers are included