search

mystic123 / tensorflow-yolo-v3

Implementation of YOLO v3 object detector in Tensorflow (TF-Slim)
https://medium.com/@pawekapica_31302/implementing-yolo-v3-in-tensorflow-tf-slim-c3c55ff59dbe
Apache License 2.0
890 stars 353 forks source link

incorrect input bounding box #110

Closed LerysG closed 2 years ago

LerysG commented 2 years ago

Hi everyone, I am struggling with yolo v5, I want to detect squares in some pics but it turns that the model does not draw the correct bounding boxes, I double checked and bounding boxes in input seems correct :

Input train image image

Bounding boxes (normalized) 0 | 0.283855 | 0.294459 | 0.087315 | 0.069557 0 | 0.485453 | 0.295452 | 0.046891 | 0.041734 0 | 0.694357 | 0.301801 | 0.066295 | 0.051009 0 | 0.282103 | 0.488188 | 0.033956 | 0.035551 0 | 0.491104 | 0.497482 | 0.054976 | 0.061829 0 | 0.686505 | 0.487019 | 0.069529 | 0.063374 0 | 0.285435 | 0.683627 | 0.063061 | 0.058737 0 | 0.485946 | 0.677126 | 0.080847 | 0.081923 0 | 0.695194 | 0.688837 | 0.084081 | 0.081923

Image in the model after resizing (under runs/train/exp) image

Anyone has a clue? Many thanks in advances. Yolo v5 repo pulled on 08 feb 2022.

LerysG commented 2 years ago

I found the issue: one has to delete the cache files.