search

chuanqi305 / MobileNet-SSD

Caffe implementation of Google MobileNet SSD detection network, with pretrained weights on VOC0712 and mAP=0.727.
MIT License
2k stars 1.18k forks source link

Training aborts when running train.sh for unknown reason #187

Closed tahriribraq closed 3 years ago

tahriribraq commented 3 years ago

Hello, I have carefully followed all your instructions and was able to successfully generate lmdbs for my dataset. I'm using images from the voc dataset as well as some of my own, and I added one more class to the 21 present in the voc labelmap, so 22 in total. I was also able to generate the prototxt files using gen_model.sh and passing in 22 as classnum. However, as you can see in the attachment, when running train.sh the training is aborted after creating layer conv13. I'm using cpu-only caffe and running this on Ubuntu 20.04 in a 30GB linux partitioned disk on my laptop and 13GB of free space is available. I've done all the troubleshooting I possibly could and totally clueless why the training aborts. Please help and let me know if you need any further info train_error :(

tahriribraq commented 3 years ago

Thankfully I was able to resolve this issue by referring to another similar issue: https://github.com/chuanqi305/MobileNet-SSD/issues/180