Closed fl82hope closed 2 years ago
The provided config https://github.com/open-mmlab/mmdetection/blob/master/configs/wider_face/ssd300_wider_face.py, the training loss is always Nan, Any solutions?
DSFD (https://github.com/Tencent/FaceDetection-DSFD) is a practical face detector, any plan to incorporate it into mmdetection?
For the second question, we mainly support general object detection, so we have not plan to reimplement it. For the first question, how many GPUs did you use? You can set grad_clip in the optimizer_config.
grad_clip
optimizer_config
The provided config https://github.com/open-mmlab/mmdetection/blob/master/configs/wider_face/ssd300_wider_face.py, the training loss is always Nan, Any solutions?
DSFD (https://github.com/Tencent/FaceDetection-DSFD) is a practical face detector, any plan to incorporate it into mmdetection?