Add YOLOv5Face model support #38
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Collaborator
DefTruth
commented
Jul 25, 2022
•
DefTruth
commented
- add YOLOv5Face model support
- add FaceDetectionResult、Sort/NMS/Visualize methods for FaceDetectionResult
- format ScaledYOLOv4
- fixed runtime_option print func bugs
merge from paddlepaddle/fastdeploy:develop
jiangjiajun
approved these changes
Jul 25, 2022
| std::vector<std::array<float, 4>> boxes; | ||
| // landmark: x, y, landmarks may empty if the | ||
| // model don't detect face with landmarks. | ||
| // Note, one face might have multiple landmarks, |
There was a problem hiding this comment.
什么类型的模型会没有landmark； 当没有landmark时，模型的开发应该是用DetectionResullt，还是用FaceDetectionResult呢
There was a problem hiding this comment.
- 关于什么时候会没有landmark？
在人脸检测中，带landmark和不带landmark的模型都是非常常见的。landmark通常作为一个辅助任务出现在人脸检测的算法中。所以定义一个FaceDetectionResult确实是需要考虑这两种情况。以下是一些例子：
- 带landmark的人脸检测器：YOLOv5Face、RetinaFace、SCRFD、CenterFace、BlazeFace等等
- 不带landmark的人脸检测器：DSFD(Tencent)、SSH、UltraFace、LFFD、LFD、S3FD等等
- 关于若不带landmark是否复用DetectionResullt？
个人观点：不建议。理由有以下3个：
- 首先，人脸检测和通用目标检测确实应该是两个不同的cv子领域，虽然说有一定的相似性，但从任务上来说是不一样的，都用DetectionResullt感觉有点奇怪
- 人脸检测没有多分类的label_id，只有人脸，目标检测通常是多类别的，需要有label_id；因此人脸检测也不需要像yolov5那样在做NMS时先通过label_id*max_wh来做offset提高recall
- 很多的人脸检测器带有landmark，但是大部分目标检测器是不带landmark的
# for free
to join this conversation on GitHub.
Already have an account?
# to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.