modify tensforflow plugin syntax
Moe
parent
766454becb
commit
0b7a094506
|
|
@ -1,4 +1,9 @@
|
|||
#!/bin/bash
|
||||
mkdir data
|
||||
mkdir data/inception
|
||||
chmod -R 777 data
|
||||
wget https://cdn.shinobi.video/weights/inception5h.zip -O inception5h.zip
|
||||
unzip inception5h.zip -d data/inception
|
||||
if [ $(dpkg-query -W -f='${Status}' opencv_version 2>/dev/null | grep -c "ok installed") -eq 0 ]; then
|
||||
echo "Shinobi - Do ypu want to let the `opencv4nodejs` npm package install OpenCV? "
|
||||
echo "Only do this if you do not have OpenCV already or will not use a GPU (Hardware Acceleration)."
|
||||
|
|
|
|||
|
|
@ -1,212 +0,0 @@
|
|||
//
|
||||
// Shinobi - OpenCV Plugin
|
||||
// Copyright (C) 2016-2025 Moe Alam, moeiscool
|
||||
//
|
||||
// # Donate
|
||||
//
|
||||
// If you like what I am doing here and want me to continue please consider donating :)
|
||||
// PayPal : paypal@m03.a
|
||||
//
|
||||
process.on('uncaughtException', function (err) {
|
||||
console.error('uncaughtException',err);
|
||||
});
|
||||
var fs=require('fs');
|
||||
var cv=require('opencv');
|
||||
var exec = require('child_process').exec;
|
||||
var moment = require('moment');
|
||||
var config=require('./conf.json');
|
||||
if(config.systemLog===undefined){config.systemLog=true}
|
||||
s={
|
||||
group:{},
|
||||
dir:{
|
||||
cascades:__dirname+'/cascades/'
|
||||
},
|
||||
isWin:(process.platform==='win32')
|
||||
}
|
||||
//default stream folder check
|
||||
if(!config.streamDir){
|
||||
if(s.isWin===false){
|
||||
config.streamDir='/dev/shm'
|
||||
}else{
|
||||
config.streamDir=config.windowsTempDir
|
||||
}
|
||||
if(!fs.existsSync(config.streamDir)){
|
||||
config.streamDir=__dirname+'/streams/'
|
||||
}else{
|
||||
config.streamDir+='/streams/'
|
||||
}
|
||||
}
|
||||
s.dir.streams=config.streamDir;
|
||||
//streams dir
|
||||
if(!fs.existsSync(s.dir.streams)){
|
||||
fs.mkdirSync(s.dir.streams);
|
||||
}
|
||||
s.gid=function(x){
|
||||
if(!x){x=10};var t = "";var p = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789";
|
||||
for( var i=0; i < x; i++ )
|
||||
t += p.charAt(Math.floor(Math.random() * p.length));
|
||||
return t;
|
||||
};
|
||||
s.systemLog=function(q,w,e){
|
||||
if(!w){w=''}
|
||||
if(!e){e=''}
|
||||
if(config.systemLog===true){
|
||||
return console.log(moment().format(),q,w,e)
|
||||
}
|
||||
}
|
||||
s.findCascades=function(callback){
|
||||
var tmp={};
|
||||
tmp.foundCascades=[];
|
||||
fs.readdir(s.dir.cascades,function(err,files){
|
||||
files.forEach(function(cascade,n){
|
||||
if(cascade.indexOf('.xml')>-1){
|
||||
tmp.foundCascades.push(cascade.replace('.xml',''))
|
||||
}
|
||||
})
|
||||
s.cascadesInDir=tmp.foundCascades;
|
||||
callback(tmp.foundCascades)
|
||||
})
|
||||
}
|
||||
s.findCascades(function(){
|
||||
//get cascades
|
||||
})
|
||||
s.detectObject=function(buffer,d){
|
||||
var keys = Object.keys(d.mon.detector_cascades);
|
||||
if(d.mon.detector_lisence_plate==="1"){
|
||||
if(!d.mon.detector_lisence_plate_country||d.mon.detector_lisence_plate_country===''){
|
||||
d.mon.detector_lisence_plate_country='us'
|
||||
}
|
||||
d.tmpFile=s.gid(5)+'.jpg'
|
||||
if(!fs.existsSync(s.dir.streams)){
|
||||
fs.mkdirSync(s.dir.streams);
|
||||
}
|
||||
d.dir=s.dir.streams+d.ke+'/'
|
||||
if(!fs.existsSync(d.dir)){
|
||||
fs.mkdirSync(d.dir);
|
||||
}
|
||||
d.dir=s.dir.streams+d.ke+'/'+d.id+'/'
|
||||
if(!fs.existsSync(d.dir)){
|
||||
fs.mkdirSync(d.dir);
|
||||
}
|
||||
fs.writeFile(d.dir+d.tmpFile,buffer,function(err){
|
||||
if(err) return s.systemLog(err);
|
||||
exec('alpr -j -c '+d.mon.detector_lisence_plate_country+' '+d.dir+d.tmpFile,{encoding:'utf8'},(err, scan, stderr) => {
|
||||
if(err){
|
||||
s.systemLog(err);
|
||||
}else{
|
||||
try{
|
||||
try{
|
||||
scan=JSON.parse(scan)
|
||||
}catch(err){
|
||||
if(!scan||!scan.results){
|
||||
return s.systemLog(scan,err);
|
||||
}
|
||||
}
|
||||
if(scan.results.length>0){
|
||||
scan.plates=[]
|
||||
scan.mats=[]
|
||||
scan.results.forEach(function(v){
|
||||
v.candidates.forEach(function(g,n){
|
||||
if(v.candidates[n].matches_template)
|
||||
delete(v.candidates[n].matches_template)
|
||||
})
|
||||
scan.plates.push({coordinates:v.coordinates,candidates:v.candidates,confidence:v.confidence,plate:v.plate})
|
||||
var width = Math.sqrt( Math.pow(v.coordinates[1].x - v.coordinates[0].x, 2) + Math.pow(v.coordinates[1].y - v.coordinates[0].y, 2));
|
||||
var height = Math.sqrt( Math.pow(v.coordinates[2].x - v.coordinates[1].x, 2) + Math.pow(v.coordinates[2].y - v.coordinates[1].y, 2))
|
||||
scan.mats.push({
|
||||
x:v.coordinates[0].x,
|
||||
y:v.coordinates[0].y,
|
||||
width:width,
|
||||
height:height,
|
||||
tag:v.plate
|
||||
})
|
||||
})
|
||||
tx({f:'trigger',id:d.id,ke:d.ke,details:{plug:config.plug,name:'licensePlate',reason:'object',matrices:scan.mats,confidence:d.average,imgHeight:d.mon.detector_scale_y,imgWidth:d.mon.detector_scale_x,frame:d.base64}})
|
||||
}
|
||||
}catch(err){
|
||||
s.systemLog(err);
|
||||
}
|
||||
}
|
||||
exec('rm -rf '+d.dir+d.tmpFile,{encoding:'utf8'})
|
||||
})
|
||||
})
|
||||
}
|
||||
if(keys.length===0){return false}
|
||||
cv.readImage(buffer, function(err,im){
|
||||
if(err){console.log(err);return false;}
|
||||
var width = im.width();
|
||||
var height = im.height();
|
||||
|
||||
if (width < 1 || height < 1) {
|
||||
throw new Error('Image has no size');
|
||||
}
|
||||
keys.forEach(function(v,n){
|
||||
im.detectObject(s.dir.cascades+v+'.xml',{}, function(err,mats){
|
||||
if(err){console.log(err);return false;}
|
||||
if(mats&&mats.length>0){
|
||||
s.cx({f:'trigger',id:d.id,ke:d.ke,details:{plug:config.plug,name:v,reason:'object',matrices:mats,confidence:d.average,imgHeight:height,imgWidth:width}})
|
||||
}
|
||||
})
|
||||
})
|
||||
})
|
||||
}
|
||||
io = require('socket.io-client')('ws://'+config.host+':'+config.port);//connect to master
|
||||
s.cx=function(x){x.pluginKey=config.key;x.plug=config.plug;return io.emit('ocv',x)}
|
||||
io.on('connect',function(d){
|
||||
s.cx({f:'init',plug:config.plug});
|
||||
})
|
||||
io.on('disconnect',function(d){
|
||||
io.connect()
|
||||
})
|
||||
io.on('f',function(d){
|
||||
switch(d.f){
|
||||
case'refreshPlugins':
|
||||
s.findCascades(function(cascades){
|
||||
s.cx({f:'s.tx',data:{f:'detector_cascade_list',cascades:cascades},to:'GRP_'+d.ke})
|
||||
})
|
||||
break;
|
||||
case'readPlugins':
|
||||
s.cx({f:'s.tx',data:{f:'detector_cascade_list',cascades:s.cascadesInDir},to:'GRP_'+d.ke})
|
||||
break;
|
||||
case'init_monitor':
|
||||
if(s.group[d.ke]&&s.group[d.ke][d.id]){
|
||||
s.group[d.ke][d.id].canvas={}
|
||||
s.group[d.ke][d.id].canvasContext={}
|
||||
s.group[d.ke][d.id].blendRegion={}
|
||||
s.group[d.ke][d.id].blendRegionContext={}
|
||||
s.group[d.ke][d.id].lastRegionImageData={}
|
||||
delete(s.group[d.ke][d.id].cords)
|
||||
delete(s.group[d.ke][d.id].buffer)
|
||||
}
|
||||
break;
|
||||
case'frame':
|
||||
d.details={}
|
||||
try{
|
||||
if(!s.group[d.ke]){
|
||||
s.group[d.ke]={}
|
||||
}
|
||||
if(!s.group[d.ke][d.id]){
|
||||
s.group[d.ke][d.id]={
|
||||
canvas:{},
|
||||
canvasContext:{},
|
||||
lastRegionImageData:{},
|
||||
blendRegion:{},
|
||||
blendRegionContext:{},
|
||||
}
|
||||
}
|
||||
if(!s.group[d.ke][d.id].buffer){
|
||||
s.group[d.ke][d.id].buffer=[d.frame];
|
||||
}else{
|
||||
s.group[d.ke][d.id].buffer.push(d.frame)
|
||||
}
|
||||
if(d.frame[d.frame.length-2] === 0xFF && d.frame[d.frame.length-1] === 0xD9){
|
||||
s.group[d.ke][d.id].buffer=Buffer.concat(s.group[d.ke][d.id].buffer);
|
||||
s.detectObject(s.group[d.ke][d.id].buffer,d)
|
||||
s.group[d.ke][d.id].buffer=null;
|
||||
}
|
||||
} catch(err){
|
||||
console.error(err)
|
||||
}
|
||||
break;
|
||||
}
|
||||
})
|
||||
|
|
@ -16,6 +16,7 @@ var exec = require('child_process').exec;
|
|||
var moment = require('moment');
|
||||
var Canvas = require('canvas');
|
||||
var express = require('express');
|
||||
const path = require('path');
|
||||
var http = require('http'),
|
||||
app = express(),
|
||||
server = http.createServer(app);
|
||||
|
|
@ -135,93 +136,91 @@ s.detectLicensePlate=function(buffer,d,tx){
|
|||
})
|
||||
}
|
||||
s.detectObject=function(buffer,d,tx){
|
||||
//detect license plate?
|
||||
if(d.mon.detector_lisence_plate==="1"){
|
||||
s.detectLicensePlate(buffer,d,tx)
|
||||
}
|
||||
cv.imdecodeAsync(buffer,(err,im) => {
|
||||
if(err){
|
||||
console.log(err)
|
||||
return
|
||||
}
|
||||
const cv = require('../');
|
||||
const fs = require('fs');
|
||||
const path = require('path');
|
||||
//detect license plate?
|
||||
if(d.mon.detector_lisence_plate==="1"){
|
||||
s.detectLicensePlate(buffer,d,tx)
|
||||
}
|
||||
cv.imdecodeAsync(buffer,(err,im) => {
|
||||
if(err){
|
||||
console.log(err)
|
||||
return
|
||||
}
|
||||
|
||||
if (!cv.xmodules.dnn) {
|
||||
throw new Error('exiting: opencv4nodejs compiled without dnn module');
|
||||
}
|
||||
if (!cv.xmodules.dnn) {
|
||||
throw new Error('exiting: opencv4nodejs compiled without dnn module');
|
||||
}
|
||||
|
||||
// replace with path where you unzipped inception model
|
||||
const inceptionModelPath = '__dirname/tf-inception';
|
||||
// replace with path where you unzipped inception model
|
||||
const inceptionModelPath = __dirname+'/data/inception';
|
||||
|
||||
|
||||
const modelFile = path.resolve(inceptionModelPath, 'tensorflow_inception_graph.pb');
|
||||
const classNamesFile = path.resolve(inceptionModelPath, 'imagenet_comp_graph_label_strings.txt');
|
||||
if (!fs.existsSync(modelFile) || !fs.existsSync(classNamesFile)) {
|
||||
console.log('could not find inception model');
|
||||
console.log('download the model from: https://storage.googleapis.com/download.tensorflow.org/models/inception5h.zip');
|
||||
throw new Error('exiting');
|
||||
}
|
||||
const modelFile = path.resolve(inceptionModelPath, 'tensorflow_inception_graph.pb');
|
||||
const classNamesFile = path.resolve(inceptionModelPath, 'imagenet_comp_graph_label_strings.txt');
|
||||
if (!fs.existsSync(modelFile) || !fs.existsSync(classNamesFile)) {
|
||||
console.log('could not find inception model');
|
||||
console.log('download the model from: https://cdn.shinobi.video/weights/inception5h.zip');
|
||||
throw new Error('exiting');
|
||||
}
|
||||
|
||||
// read classNames and store them in an array
|
||||
const classNames = fs.readFileSync(classNamesFile).toString().split('\n');
|
||||
// read classNames and store them in an array
|
||||
const classNames = fs.readFileSync(classNamesFile).toString().split('\n');
|
||||
|
||||
// initialize tensorflow inception model from modelFile
|
||||
const net = cv.readNetFromTensorflow(modelFile);
|
||||
// initialize tensorflow inception model from modelFile
|
||||
const net = cv.readNetFromTensorflow(modelFile);
|
||||
|
||||
const classifyImg = (im) => {
|
||||
// inception model works with 224 x 224 images, so we resize
|
||||
// our input images and pad the image with white pixels to
|
||||
// make the images have the same width and height
|
||||
const maxImgDim = 224;
|
||||
const white = new cv.Vec(255, 255, 255);
|
||||
const imgResized = img.resizeToMax(maxImgDim).padToSquare(white);
|
||||
// inception model works with 224 x 224 images, so we resize
|
||||
// our input images and pad the image with white pixels to
|
||||
// make the images have the same width and height
|
||||
const maxImgDim = 224;
|
||||
const white = new cv.Vec(255, 255, 255);
|
||||
const imgResized = im.resizeToMax(maxImgDim).padToSquare(white);
|
||||
|
||||
// network accepts blobs as input
|
||||
const inputBlob = cv.blobFromImage(imgResized);
|
||||
net.setInput(inputBlob);
|
||||
// network accepts blobs as input
|
||||
const inputBlob = cv.blobFromImage(imgResized);
|
||||
net.setInput(inputBlob);
|
||||
|
||||
// forward pass input through entire network, will return
|
||||
// classification result as 1xN Mat with confidences of each class
|
||||
const outputBlob = net.forward();
|
||||
// forward pass input through entire network, will return
|
||||
// classification result as 1xN Mat with confidences of each class
|
||||
const outputBlob = net.forward();
|
||||
|
||||
// find all labels with a minimum confidence
|
||||
const minConfidence = 0.05;
|
||||
const locations =
|
||||
outputBlob
|
||||
.threshold(minConfidence, 1, cv.THRESH_BINARY)
|
||||
.convertTo(cv.CV_8U)
|
||||
.findNonZero();
|
||||
|
||||
const result =
|
||||
locations.map(pt => ({
|
||||
confidence: parseInt(outputBlob.at(0, pt.x) * 100) / 100,
|
||||
className: classNames[pt.x]
|
||||
}))
|
||||
// sort result by confidence
|
||||
.sort((r0, r1) => r1.confidence - r0.confidence)
|
||||
.map(res => `${res.className} (${res.confidence})`);
|
||||
if(result.length > 0) {
|
||||
s.cx({
|
||||
f:'trigger',
|
||||
id:d.id,
|
||||
ke:d.ke,
|
||||
name:cascade,
|
||||
details:{
|
||||
plug:'tensorflow',
|
||||
name:'tensorflow',
|
||||
reason:result,
|
||||
matrices : matrices
|
||||
// confidence:d.average
|
||||
},
|
||||
imgHeight:d.mon.detector_scale_y,
|
||||
imgWidth:d.mon.detector_scale_x
|
||||
})
|
||||
}
|
||||
// return result;
|
||||
};
|
||||
})
|
||||
// find all labels with a minimum confidence
|
||||
const minConfidence = 0.05;
|
||||
const locations =
|
||||
outputBlob
|
||||
.threshold(minConfidence, 1, cv.THRESH_BINARY)
|
||||
.convertTo(cv.CV_8U)
|
||||
.findNonZero();
|
||||
// locations.forEach(function(v){
|
||||
// console.log(v)
|
||||
// })
|
||||
const result =
|
||||
locations.map(pt => ({
|
||||
confidence: parseInt(outputBlob.at(0, pt.x) * 100) / 100,
|
||||
className: classNames[pt.x]
|
||||
}))
|
||||
// sort result by confidence
|
||||
.sort((r0, r1) => r1.confidence - r0.confidence)
|
||||
.map(res => `${res.className} (${res.confidence})`);
|
||||
console.log(result)
|
||||
if(result.length > 0) {
|
||||
s.cx({
|
||||
f:'trigger',
|
||||
id:d.id,
|
||||
ke:d.ke,
|
||||
name:'tensorflow',
|
||||
details:{
|
||||
plug:'tensorflow',
|
||||
name:'tensorflow',
|
||||
reason:'object',
|
||||
matrices : result
|
||||
// confidence:d.average
|
||||
},
|
||||
imgHeight:d.mon.detector_scale_y,
|
||||
imgWidth:d.mon.detector_scale_x
|
||||
})
|
||||
}
|
||||
})
|
||||
}
|
||||
s.systemLog=function(q,w,e){
|
||||
if(!w){w=''}
|
||||
if(!e){e=''}
|
||||
|
|
|
|||
Loading…
Reference in New Issue