toandaominh1997 / EfficientDet.Pytorch

Implementation EfficientDet: Scalable and Efficient Object Detection in PyTorch
MIT License
1.44k stars 305 forks source link

No boxes to NMS #143

Open SaiAinala3 opened 4 years ago

SaiAinala3 commented 4 years ago

Hi,

I have trained using : python3 train.py --dataset COCO --dataset_root data/coco/ --network efficientdet-d0 --batch_size 8 --num_epoch 15 --num_class 1 --save_folder saved/

After 5 epochs:

cls_loss       : 2.9468758384609828e-06
reg_loss       : 0.0
mean_loss      : 2.9462864633387655e-06
time           : 474.67488193511963
loss           : 2.9462864633387655e-06

15 epoch: start validation.... No boxes to NMS No boxes to NMS No boxes to NMS

Even if I run: python3 eval.py --dataset COCO --dataset_root data/coco --weight saved3/COCO/efficientdet-d0/checkpoint_14.pth

Output :

Loading checkpoint: saved3/COCO/efficientdet-d0/checkpoint_14.pth ... Loaded pretrained weights for efficientnet-b0 loading annotations into memory... Done (t=0.03s) creating index... index created! No boxes to NMS No boxes to NMS No boxes to NMS

Can someone help me with this? Thank you.