-
## error log | 日志或报错信息 | ログ
## context | 编译/运行环境 | バックグラウンド
## how to reproduce | 复现步骤 | 再現方法
1. Code and models avaible from https://github.com/FeiGeChuanShu/ncnn-Android-mediapipe_hand and co…
-
Hi, guys
I want to use the tencent ncnn library to run YOLObile in the android mobile can you provide me a trained onnx file? or tell me how to export onnx file?
Thanks
-
编译系统: win10
android_ndk: 20.1.5948944
我在win10上可以成功build ncnn, 当我想在安卓设备上运行"./benchncnn 8 4 2 -1 1"时,我需要build一个安卓版本的,于是我按照[官方教程](https://github.com/Tencent/ncnn/wiki/how-to-build#build-for-android)操作,…
-
在使用yolov5/yolov7通过ncnn部署到android端时,发现在电脑上使用yolov5/yolov7自带的detect能够正确识别。但通过onnx、pnnx转换之后用ncnn加载的模型文件,有少部分图片识别错误的情况。感觉上是精度有一定的损失,一直困扰哪个环节出现了问题。请前辈们不吝指导,谢谢~~~
-
up主,你好,我尝试按步骤编译这个例程,但却一直报这个错误:
Cause: executing external native build for cmake /home/xxx/ncnn-android-mobilenetssd/app/src/main/jni/CMakeLists.txt
我用的NDK版本是r16b,下载的ncnn-android-vulkan-lib.zip是最新的…
-
### Describe your suggested feature
Often Manga scanlations have a lower resolution than modern phone and tablet screens, which means they will get scaled up while displaying them. This results in …
radry updated
1 month ago
-
## error log | 日志或报错信息 | ログ
Sometimes you can see SIGILL in Android log.
Sometimes device became unresponsive and need to get rebooted.
## context | 编译/运行环境 | バックグラウンド
Android arm64-v8a, SDK NDK…
-
各位大神,我在PC端ncnn下使用efficientNet时,推理正常。转到android上,推理分类完全不正确。我在pc和android上对比了中间的mat数组,在前半部分的Mat能匹配上,中间部分的Mat基本不能匹配,但正负可以匹配,并且该大的大,该小的小,相差一些值。到后面部分,直接就失真了。这是累积误差造成的吗?还是实现出现的问题。求大神指点先
-
大神好,我想用自己訓練的yolov4-tiny 在 android-ncnn 檢測目標
以下是我的yolov4-tiny-opt.param
![螢幕快照 2021-01-22 下午7 16 01](https://user-images.githubusercontent.com/41318292/105494958-7f287480-5cf6-11eb-98fa-c077d247163b.…
-
pt转onnx用到 --train
onnx->ncnn用的是pnnx
android端cpp代码,修改了in0,out0,out1,out2,generate_proposals()方法
后附param,bin,cpp(jni)文件,在一个zip里
[assets_and_cpp.zip](https://github.com/Tencent/ncnn/files/9651166/ass…