Jian comments

Results 39 comments of


                                            Jian

How to train the model??

@skyneta We are planing new face detection methods. Maybe，training code is released with next version

RuntimeError: CUDA out of memory.

Your pytorch is 1.10, our pytorch is 0.3. If you just want to do evaluation, add torch.set_grad_enabled(False) inside test_oneimage() function (at the beginning of the function).

Does anyone have other pre-trained weights instead of ResNet-151?

A 3M model is provided by https://github.com/lijiannuist/lightDSFD

Lighweight Face Detector ?

You can find lite DSFD by https://github.com/lijiannuist/lightDSFD

i got memory error when i run demo.py

Hi, demo.py adopts ResNet152 backbone and multi-scale test, which consumes much GPU memory. You can try to use less image scales for inference. We will release a more fine model.

RuntimeError: CUDA out of memory

Hi Try to use with torch.no_grad(): to inference. You can refer to https://github.com/vlad3996/FaceDetection-DSFD

请问检测一张图片的时间和mtcnn对比有提升嘛，我们做跟踪，检测用的mtcnn，但时间有点慢。

你好，DSFD的初衷是极致的人脸检测精度，因此注重使用更大的backbone和多尺度测试，这会导致速度变慢很多。

Head Detection

Preparing dataset with head bounding box label

can i use max-in-out while using sigmoid focal loss?

of course! just try it!

Is it possible to use smaller GPU for inference?

Hi @AmitRozner Of course. But Training with smaller batch size may reduce detection performance slightly.