All generated images donot have 36 bounding boxes

airsplay / py-bottom-up-attention

PyTorch bottom-up attention with Detectron2

Apache License 2.0

229 stars 56 forks source link

All generated images donot have 36 bounding boxes #21

Closed debjitpaul452 closed 3 years ago

debjitpaul452 commented 3 years ago

When I tried to generate features from my own image dataset around 60 percent of images produced 36 bounding boxes and rest produced less than 36. What is the reason?

o0t1ng0o commented 1 year ago

@debjitpaul452 Hi,

How do you solve it?