オッサンはDesktopが好き

自作PCや機械学習、自転車のことを脈絡無く書きます

Deep learning

iOSで自作U-Net

こんにちは.changです. 前回に続いてiOSでDeep Learningをやってみます. サンプルをベースに,自作U-Netの学習済みモデルを使ってSemantic Segmationします. 1. サンプルでSemantic Segmentation 2. 自作U-Netのtflite変換 3. 自作U-NetをiPhone内で推論…

iOSでTensorflowLite

こんにちは.changです. ほぼ自分用のメモですが,MacでTensorflowLiteのサンプルをビルドして,iPhoneカメラでobject detectionするまでの記録になります. 1. 必要ツールのインストール (1) Visual Studio Code (2) markdown (3) git (4) rbenv (5) Cocoa…

超今更だけど、kerasの旧ver.でマルチクラスのU-Netが書けない件

こんにちは.changです. kerasでマルチクラスのU-Netを書くとちょっと変だな,と以前から気になっていました. 今回,それが解決したので記録します. 結果だけを言うと,ver.2になって不具合(?)が治ったという話です. 1. Tensorflow ver.1 + kerasによるU…

C code + CPU is the best choice for Tensorflow inference in general purpose Windows

This article is a translation of Japanese ver. Original ver. is here*1. Hi, this is chang. Today, I wrote my personal opinion about how to implement Tensorflow for inference into general purpose Windows computers. To be honest, I wanted to…

Dig data does not make AI smarter

This article is a translation of Japanese ver. 日本語版はこちら*1 Hi, this is chang. Today, I introduce my personal opinion about AI: Learning big data does not make AI smarter. 0. What is Big data? Wikipedia shows that "big data is a fiel…

Accelerate inference by running tensorflow's session in advance

This article is a translation of Japanese ver. 日本語版はこちら*1 Hi, this is chang. Today, I tried to accelerate inference. I published defection viewer not long ago*2*3. Its inference took about 6 sec. I investigated why it took so long …

Deep Learning: Rotation invariance and multi-kind defection with different size

This article is a translation of Japanese ver. 日本語版はこちら*1 Hi, this is chang. Recently, I introduced multi-kind defection using U-Net*2. After the publication, I noticed that the detection of Class_2 was bad and made an investigatio…

Deep learning: multi-channel is very useful for multi-kind defection

This article is a translation of Japanese ver. 日本語版はこちら*1 Hi, this is chang. Previously, I tried to make 6 Channel U-Net learn defects with multiple kinds. Then I wrote "1 channel network for each kind seemed to be better"*2. Toda…

Deep Learning using only one defected image

Note: this article is a translation of Japanese ver. 日本語版はこちら*1 Hi, this is chang. Today I tried to consider how to construct deep learning using the small number of defected images. 0. Background Previously, I introduced the defec…