-
paddle版本:develop
预测脚本数据和模型:
[infer_blazenas_mkldnn.zip](https://github.com/PaddlePaddle/Paddle/files/3803772/infer_blazenas_mkldnn.zip)
报错信息:
![Uploading 屏幕快照 2019-11-04 下午7.35.36.png…](…
ceci3 updated
4 years ago
-
convert pytorch to onnx
```python
# -*- coding: utf-8 -*-
import torchvision.models as models
import torch
import torch.nn as nn
mobilenet = models.resnet18()
# dummy_input = torch.te…
-
您好,这边有两个疑问:
一是face_detection的infer只用det0 = detect_face(image, shrink),fps=5左右,是否可以提高
二是face_detection的infer目前只支持单张图像的infer,是否可以设置batchsize=10这样
-
首先感谢分享这个代码,很简洁,效果很好。关于现有模型我有几个问题
- 用 blazeface 模型,优化速度
- 300VW 视频人脸数据集
- 模型只用正方形框来预测人脸,anchor 长宽比只设置为1:1
据说blazeface的模型速度很快,打算在这个基础上测试以下,不知测试过没有
另外,在移动端,人脸检测大多用于视频情况,这个时候检测框的抖动会非常明显,一般有两个解决思路,1…
-
你好,首先感谢你的代码;我按照你所说的做法,已经基本跑通了模型。但是我在训练的时候,loss一直在10左右震荡,并没有看到收敛的趋势,请问这样子可能是哪里出了问题呢?你最后训练得到的loss是多少呢?
如果方便的话,不知道可不可以上传一下你训练的模型(基于tensorflow的BlazeFace)
谢谢
-
well when I try to detect faces using `face_detector.detect_from_batch` , output bboxes differs according to video (frames size)
for example
here is the two videos in different sizes
original : …
-
Hi
I studied this paper `BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs`, which led me to download `face_detection_front.tflite`.
But their model is completely different. I read…
-
## 🐛 Bug
In Pytorch 1.2.0, a new 'debug' folder with a fairly large pickled file has come in to existence.
I have a 3MB model and this debug folder seems to take up an extra 1MB making it a 4MB mo…
-
I want to try the blazeface model using back camera, but I only find face_detection_front.tflite. Do you guys have face_detection_back.tflite? Thank you.
-