search

Borda / keras-yolo3

A Keras implementation of YOLOv3 (Tensorflow backend) a successor of qqwweee/keras-yolo3
MIT License
31 stars 10 forks source link

Q: IOU by kmeans so small? #22

Closed jwnirvana closed 5 years ago

jwnirvana commented 5 years ago

NICE job. But why is iou by kmeans so small (4%)?

Borda commented 5 years ago

thx, could you elaborate a bit more what you mean by "iou by kmeans so small (4%)"?

jwnirvana commented 5 years ago

thx, could you elaborate a bit more what you mean by "iou by kmeans so small (4%)"?

Have you run the scripts/kmeans.py by yourself? How about your result?

jwnirvana commented 5 years ago

thx, could you elaborate a bit more what you mean by "iou by kmeans so small (4%)"?

in ./scripts/kmeans.py: logging.info('Accuracy: %f', self.avg_iou(all_boxes, result) * 100) the avg_iou printed is so small

Borda commented 5 years ago

I am afraid that from this information I am not able to reproduce your issue... I just finished update of clustering.py and for my dataset, it gives me 82%

Borda commented 5 years ago

@jwnirvana I believe that it is solved for now, but if you have any further question, feel free to reopen this issue...