vlad3996

Use with `torch.no_grad():` to inference. You can also try my simple demo.py with original author's checkpoint https://github.com/vlad3996/FaceDetection-DSFD

You can use this https://github.com/vlad3996/FaceDetection-DSFD with original author's checkpoint or try at least `with torch.no_grad():` to inference

You can try to use my fork : https://github.com/vlad3996/lsc-cnn

@learnermaxRL the problem is with input image shape : it must be divisible by 16 because author concatenate feature maps from 3rd pooling layer and upsampled by nn.ConvTranspose2d maps from...

You can use centers of head/face bounding boxes to obtain GT dots, which you can transform to density map via gaussian kernels.

I've obtained 10.42 MAE and 16.89 MSE on part B without augmentation but it takes about 20 hours on GTX 1080Ti. So maybe augmentation is not needed at all.

I've obtained 10.42 MAE and 16.89 MSE on part B without augmentation and with other default params from repo.

I've tried to train with higher lr, another optimizes (Adam, Adadelta, COCOB), change loss function during training process, but can't achieve any better results. Finally I've used author's SGD with...

@liuleiBUAA your change in __init__ is some kind useless : you can load weights by providing checkpoint via arg --pre : ``` python checkpoint = torch.load(args.pre) model.load_state_dict(checkpoint['state_dict']) ``` I don't...

@libaikai I've just cloned original repo and run training with python 2.7, pytorch 0.4.1. No errors. Do you use VGG16 pre-trained weights? It's a little bit tricky to download weights...