Xuan Lin

@YING1992 How did you get your data set, did you follow the instructions in README to generate training set for Rnet and Onet?

@JoeHEZHAO For bbox regression task, the gradients of negative examples will be set to 0 in backward, [https://github.com/Seanlinx/mtcnn/blob/master/core/negativemining.py#L53](https://github.com/Seanlinx/mtcnn/blob/master/core/negativemining.py#L53) so their ground truth can be any value you like. And it's...

@JoeHEZHAO Yes, you're right. I didn't follow the ratio of losses provided in the paper since my implication doesn't have landmark task included. I simply assign equal importance to cls...

You don't have a folder named "mtcnn_pnet" under ./data/cache so it fails to generate the .pkl file. It's a simple problem if you read the code where the error occurs.

@Cv9527 Not necessary, but adding mirrored images will enlarge the training set and improve the performance.

@Cv9527 The negativemining layer sets the out_grad to "0" where it should be ignored.

@hkdqliu hi, the ratio of pos:part:neg I used in training mtcnn Rnet is about 1:1:2. But I don't think the network structure of mtcnn is suitable for car detection for...

@GreenKing The log looks reasonable, I got similar result at the end of pnet. If you want to achieve similar performance, the thresholds you used to generate training data for...

@fninsiima (x1, y1), (x2, y2) are the coordinates of the left upper corner and lower right corner of bbox respectively. 'width' and 'height' are the width and height of the...

@fninsiima (x1,y1),(x2,y2) are the coordinates of the training sample in their original images. (xgt, ygt) stands for the left upper or lower right of the ground truth bbox. I just...