Open kevin29a opened 4 years ago
The big question I have is should I be able to achieve the results I see on the online demo myself? Things work great for me on that site, but I'm not able to reproduce those results locally, including the examples in the repo. Is the source code available for the server running at https://justadudewhohacks.github.io/face-api.js/face_and_landmark_detection/ ?
The big question I have is should I be able to achieve the results I see on the online demo myself? Things work great for me on that site, but I'm not able to reproduce those results locally, including the examples in the repo. Is the source code available for the server running at https://justadudewhohacks.github.io/face-api.js/face_and_landmark_detection/ ?
Agree, I have the same issues and same question.
Trying the example/demo site there is no indication of faces being detected at all.
I did some more research on this. For me, things work well on version 0.21.0. They start to break on 0.22.0. Is that the same for you?
Can you guys guide, i need api, to compare 2 images from url and send the result back instead of using it from form.
Hi, have anyone found the issue. I also facing same with multiple records in DB, when the pic only contains clear face image and not any background/shape, the accurate results coming, else wrong matches. I am using 0.22.2 in Nodejs app
I've switched to using 0.21.0. Far from an ideal solution. The commit to upgrade to 0.22.0 is quite large, and I haven't had the time to dig through there to look for the issue.
I would be curious if the downgrade fixes things for someone else as well. If there are multiple people for whom the downgrade works, it would make me more inspired to dig into the codebase myself.
Thanks for the suggestion, Kevin. I will try downgrading to that version and let you know.
On a macbook running the examples from the master branch, I get poor accuracy on face detection (see screenshots attached). Is the online demo at https://justadudewhohacks.github.io/face-api.js/face_and_landmark_detection running current code? Also would note that I get good accuracy on my linux computer working against the master branch. My linux laptop has an intel video card, and my macbook has both an nvidia and intel gpu.