search

loiccordone / object-detection-with-spiking-neural-networks

Repository code for the IJCNN 2022 paper "Object Detection with Spiking Neural Networks on Automotive Event Data"
MIT License
57 stars 12 forks source link

About pretrained model #14

Closed xxyll closed 1 year ago

xxyll commented 2 years ago

Hi, long time no see. I find a problem I hadn't noticed before. After training, there will be generate a folder ‘ckpt-od-gen1-vgg-11’, and there are three '.ckpt' files in it. After training the mAp is 1.162, the loss of epoch49 is 0.744. But the files in ‘ckpt-od-gen1-vgg-11’ is 'gen1-epoch=00-train_loss=9.3256.ckpt', 'gen1-epoch=01-train_loss=2.2496.ckpt', 'gen1-epoch=02-train_loss=1.7713.ckpt'. It seems that only the results of the first three rounds have been saved. Is the pretrained model used one of the three files in the test step？

xxyll commented 1 year ago

I solved it!! I modified 'mode='max'' to 'mode ='min'' in line 101~~