Closed hdjsjyl closed 6 years ago
@hdjsjyl Hi, thanks for your interest, you can download the pretrained model here: https://github.com/Superlee506/Mask_RCNN_Humanpose/releases
@Superlee506 , thanks for your reply. I solved this problem. Actually, I want to fit human pose estimation part to facial landmarks. Do you think it is reasonable? Can you give me some advice? Any advice will be appreciated. Thanks
@hdjsjyl For facial landmarks detection, you should firstly change the bbox head for face detecation rather than human detection. Secondly, change the keypoint head for face.
@Superlee506 , thanks for your advice. I will read your code line by line. Thank you very much.
@Superlee506 , Now I have changed segmentation part for keypoint detection. But I found that rcnnL1Loss will be very big such as 234677418896143419441152.0000 from beginning. Do you know what is the reason? Any advice will be appreciated. Thank you.
@hdjsjyl pycocotools is a third-party package for operating the coco dataset. You can install it as described in https://github.com/cocodataset/cocoapi
Hi, thanks for your code. It is very nice. I meet a problem, when I run inference_humanpose.ipynb: and I do not find the file pycocotools, any advice will be appreciated. Thank you very much.
ImportError Traceback (most recent call last)