deepinsight / deepinsight/insightface

SCRFD outputs bounding boxes and kps which are shfited to right for some faces of the insightface/sample-images/t2.jpg

Open
#1,583 0 comments 2 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
29.7k
Forks
6.1k
PR merge metrics
No merged PRs in 30d

Description

Run scrfd.py on the t2.jpg with an onnx model converted from the pytorch model (using scrfd2onnx.py), for example scrfd_2.5g_bnkps_shape640x640.onnx, the bounding boxes and kps are shifted to the right for some of the faces of the sample image t2.jpg. See the attached.

![t2_2 5g_bnkps](https://user-images.githubusercontent.com/9027847/121816031-4d815d00-cc47-11eb-932b-d4dd56aa1418.jpg)

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by running scrfd.py on insightface/sample-images/t2.jpg with scrfd_2.5g_bnkps_shape640x640.onnx, then inspect scrfd2onnx.py and compare the converted model's boxes and keypoints with the PyTorch model. Done means the affected faces in t2.jpg have correctly aligned bounding boxes and keypoints after conversion.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.