-
Notifications
You must be signed in to change notification settings - Fork 1.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
multi_pose #739
Comments
I didn't uderstand the first question. The rest are correct as far as I can see. |
hey,I am also interested in human pose estimation.Is it convenient for you to share your understanding of the parameters?Thanks in advance!!! |
Thank you for sharing this code. I would like to ask you more about_ Don't directly regress 17 corners in post as in target detection?
Another one is about multi_ The understanding of some parameters in post is right. Please correct them:
#中心点的热力图
hm = np.zeros((self.num_classes, output_res, output_res), dtype=np.float32)
#关节点的热力图
hm_hp = np.zeros((num_joints, output_res, output_res), dtype=np.float32)
#不同物体中心是否在一起的设置(代码中没用到)
dense_kps = np.zeros((num_joints, 2, output_res, output_res),
dtype=np.float32)
# 不同物体中心是否存在设置(代码中没用到)
dense_kps_mask = np.zeros((num_joints, output_res, output_res),
dtype=np.float32)
#物体长宽的设置
wh = np.zeros((self.max_objs, 2), dtype=np.float32)
#关节点位置的设置
kps = np.zeros((self.max_objs, num_joints * 2), dtype=np.float32)
#中心点下采样的误差
reg = np.zeros((self.max_objs, 2), dtype=np.float32)
#中心点的位置
ind = np.zeros((self.max_objs), dtype=np.int64)
# reg_mask回归的是有无目标,以掩码mask是否等1表示,返回self.max_objs个回归mask
# 这里相当于记载一张图片存在哪些目标,有的话对应索引设置为1,其余设置为0。
reg_mask = np.zeros((self.max_objs), dtype=np.uint8)
kps_mask = np.zeros((self.max_objs, self.num_joints * 2), dtype=np.uint8)
#关节点的误差
hp_offset = np.zeros((self.max_objs * num_joints, 2), dtype=np.float32)
# 关节点的位置
hp_ind = np.zeros((self.max_objs * num_joints), dtype=np.int64)
#关节点位置是否存在
hp_mask = np.zeros((self.max_objs * num_joints), dtype=np.int64)
The text was updated successfully, but these errors were encountered: