Elad Bar
|
19db81485c
|
//
// Shinobi - DeepStack Face Recognition Plugin
// Copyright (C) 2021 Elad Bar
//
// Base Init >>
const { spawn } = require('child_process');
const fs = require('fs');
const request = require("request");
const moment = require('moment');
const config = require('./conf.json');
let s = null;
const {
workerData
} = require('worker_threads');
const isWorker = workerData && workerData.ok === true;
const pluginBasePath = isWorker ? "pluginWorkerBase.js" : "pluginBase.js";
for(let i = 0; i < 2; i++) {
try {
s = require(`../${pluginBasePath}`)(__dirname, config);
} catch(err) {
console.log(err);
s = null;
}
}
if(s === null) {
console.log(config.plug, `Plugin start has failed. ${pluginBasePath} was not found.`);
} else {
if(!isWorker) {
const {
haltMessage,
checkStartTime,
setStartTime,
} = require('../pluginCheck.js');
if(!checkStartTime()) {
console.log(haltMessage, new Date());
s.disconnectWebSocket();
}
setStartTime();
}
}
// Base Init />>
let detectorSettings = null;
const DETECTOR_TYPE_FACE = 'face';
const DETECTOR_TYPE_OBJECT = 'object';
const FACE_UNKNOWN = 'unknown';
const DEEPSTACK_API_KEY = 'api_key';
const DETECTOR_CONFIGUTATION = {
face: {
detectEndpoint: '/vision/face/recognize',
startupEndpoint: '/vision/face/list',
key: 'userid'
},
object: {
detectEndpoint: '/vision/detection',
key: 'label'
}
}
const PROTOCOLS = {
true: "https",
false: "http"
};
const log = (logger, message) => {
logger(`${moment().format()} [${config.plug}] ${message}`);
}
const logError = (message) => {
log(console.error, message);
};
const logWarn = (message) => {
log(console.warn, message);
};
const logInfo = (message) => {
log(console.info, message);
};
const logDebug = (message) => {
log(console.debug, message);
};
const postMessage = (data) => {
const message = JSON.stringify(data);
logInfo(message);
};
const initialize = () => {
const deepStackProtocol = PROTOCOLS[config.deepStack.isSSL];
baseUrl = `${deepStackProtocol}://${config.deepStack.host}:${config.deepStack.port}/v1`;
const detectionType = config.plug.split("-")[1].toLowerCase();
const detectorConfig = DETECTOR_CONFIGUTATION[detectionType];
const detectorConfigKeys = Object.keys(detectorConfig);
detectorSettings = {
type: detectionType,
active: false,
baseUrl: baseUrl,
apiKey: config.deepStack.apiKey,
jobs: []
};
if(detectionType === DETECTOR_TYPE_FACE) {
detectorSettings["registeredPersons"] = config.persons === undefined ? [] : config.persons;
}
detectorConfigKeys.forEach(k => detectorSettings[k] = detectorConfig[k]);
const testRequestData = getFormData(detectorSettings.detectEndpoint);
request.post(testRequestData, (err, res, body) => {
try {
if(err) {
throw err;
}
const response = JSON.parse(body);
if(response.error) {
detectorSettings.active = !response.error.endsWith('endpoint not activated');
} else {
detectorSettings.active = response.success;
}
const detectorSettingsKeys = Object.keys(detectorSettings);
const pluginMessageHeader = [];
pluginMessageHeader.push(`${config.plug} loaded`);
const configMessage = detectorSettingsKeys.map(k => `${k}: ${detectorSettings[k]}`);
const fullPluginMessage = pluginMessageHeader.concat(configMessage);
const pluginMessage = fullPluginMessage.join(", ");
logInfo(pluginMessage);
if (detectorSettings.active) {
s.detectObject = detectObject;
if(detectionType === DETECTOR_TYPE_FACE) {
const requestData = getFormData(detectorSettings.startupEndpoint);
const requestTime = getCurrentTimestamp();
request.post(requestData, (errStartup, resStartup, bodyStartup) => {
if (!!resStartup) {
resStartup.duration = getDuration(requestTime);
}
onFaceListResult(errStartup, resStartup, bodyStartup);
});
}
}
} catch(ex) {
logError(`Failed to initialize ${config.plug} plugin, Error: ${ex}`)
}
});
};
const processImage = (imageB64, d, tx, frameLocation, callback) => {
if(!detectorSettings.active) {
return;
}
try{
const imageStream = fs.createReadStream(frameLocation);
const form = {
image: imageStream,
min_confidence: 0.7
};
const requestData = getFormData(detectorSettings.detectEndpoint, form);
const requestTime = getCurrentTimestamp();
request.post(requestData, (err, res, body) => {
if (!!res) {
res.duration = getDuration(requestTime);
}
onImageProcessed(d, tx, err, res, body, imageB64);
fs.unlinkSync(frameLocation);
removeJob(d.ke, d.id);
});
}catch(ex){
removeJob(d.ke, d.id);
logError(`Failed to process image, Error: ${ex}`);
if(fs.existsSync(frameLocation)) {
fs.unlinkSync(frameLocation);
}
}
callback();
};
const getJobKey = (groupId, monitorId) => {
const jobKey = `${groupId}_${monitorId}`;
return jobKey;
}
const addJob = (groupId, monitorId) => {
const jobKey = getJobKey(groupId, monitorId);
const jobExists = detectorSettings.jobs.includes(jobKey);
if(!jobExists) {
detectorSettings.jobs.push(jobKey);
}
return !jobExists;
}
const removeJob = (groupId, monitorId) => {
const jobKey = getJobKey(groupId, monitorId);
detectorSettings.jobs = detectorSettings.jobs.filter(j => j !== jobKey);
}
const detectObject = (frameBuffer, d, tx, frameLocation, callback) => {
if(!detectorSettings.active) {
return;
}
const jobCreated = addJob(d.ke, d.id);
if(!jobCreated) {
return;
}
const dirCreationOptions = {
recursive: true
};
d.dir = `${s.dir.streams}${d.ke}/${d.id}/`;
frameLocation = `${d.dir}${s.gid(5)}.jpg`;
if(!fs.existsSync(d.dir)) {
fs.mkdirSync(d.dir, dirCreationOptions);
}
fs.writeFile(frameLocation, frameBuffer, function(err) {
if(err) {
removeJob(d.ke, d.id);
return s.systemLog(err);
}
try {
const imageB64 = frameBuffer.toString('base64');
processImage(imageB64, d, tx, frameLocation, callback);
} catch(ex) {
removeJob(d.ke, d.id);
logError(`Detector failed to parse frame, Error: ${ex}`);
}
});
};
const getCurrentTimestamp = () => {
const currentTime = new Date();
const currentTimestamp = currentTime.getTime();
return currentTimestamp
};
const getDuration = (requestTime) => {
const currentTime = new Date();
const currentTimestamp = currentTime.getTime();
const duration = currentTimestamp - requestTime;
return duration;
};
const onFaceListResult = (err, res, body) => {
const duration = !!res ? res.duration : 0;
try {
const response = JSON.parse(body);
const success = response.success;
const facesArr = response.faces;
const faceStr = facesArr.join(",");
if(success) {
logInfo(`DeepStack loaded with the following faces: ${faceStr}, Response time: ${duration} ms`);
} else {
logWarn(`Failed to connect to DeepStack server, Error: ${err}, Response time: ${duration} ms`);
}
} catch(ex) {
logError(`Error while connecting to DeepStack server, Error: ${ex} | ${err}, Response time: ${duration} ms`);
}
};
const onImageProcessed = (d, tx, err, res, body, imageStream) => {
const duration = !!res ? res.duration : 0;
let objects = [];
try {
if(err) {
throw err;
}
const response = JSON.parse(body);
const success = response.success;
if(success) {
const predictions = response.predictions;
if(predictions !== null && predictions.length > 0) {
objects = predictions.map(p => getDeepStackObject(p)).filter(p => !!p);
if(objects.length === 0) {
logInfo(`Processed image for ${detectorSettings.type} on monitor ${d.id} returned no results, Response time: ${duration} ms`);
} else {
const identified = objects.filter(p => p.tag !== FACE_UNKNOWN);
const unknownCount = objects.length - identified.length;
if(unknownCount > 0) {
logInfo(`{d.id}$ detected ${unknownCount} unknown ${detectorSettings.type}s, Response time: ${duration} ms`);
}
if(identified.length > 0) {
const detectedObjectsStrArr = [];
if (detectorSettings.type === DETECTOR_TYPE_FACE) {
identified.forEach(f => detectedObjectsStrArr.push(`${f.tag} (${f.person}) [${(f.confidence * 100).toFixed(2)}]`));
} else {
identified.forEach(f => detectedObjectsStrArr.push(`${f.tag} [${(f.confidence * 100).toFixed(2)}]`));
}
const detectedObjectsStr = detectedObjectsStrArr.join(", ");
logInfo(`${d.id} detected ${detectorSettings.type}s: ${detectedObjectsStr}, Response time: ${duration} ms`);
}
const isObjectDetectionSeparate = d.mon.detector_pam === '1' && d.mon.detector_use_detect_object === '1';
const width = parseFloat(isObjectDetectionSeparate && d.mon.detector_scale_y_object ? d.mon.detector_scale_y_object : d.mon.detector_scale_y);
const height = parseFloat(isObjectDetectionSeparate && d.mon.detector_scale_x_object ? d.mon.detector_scale_x_object : d.mon.detector_scale_x);
const eventData = {
f: 'trigger',
id: d.id,
ke: d.ke,
details: {
plug: config.plug,
name: d.id,
reason: detectorSettings.type,
matrices: objects,
imgHeight: width,
imgWidth: height,
time: duration,
imageStream: imageStream
}
};
tx(eventData);
}
}
} else {
logWarn(`Processed image for ${detectorSettings.type} on monitor ${d.id} failed, Reason: ${response.error}, Response time: ${duration} ms`);
}
} catch(ex) {
logError(`Error while processing image, Error: ${ex} | ${err}, Response time: ${duration} ms, Body: ${body}`);
}
return objects
};
const getFormData = (endpoint, additionalParameters) => {
const formData = {};
if(detectorSettings.apiKey) {
formData[DEEPSTACK_API_KEY] = detectorSettings.apiKey;
}
if(additionalParameters !== undefined && additionalParameters !== null) {
const keys = Object.keys(additionalParameters);
keys.forEach(k => formData[k] = additionalParameters[k]);
}
const requestData = {
url: `${detectorSettings.baseUrl}${endpoint}`,
time: true,
formData: formData
};
return requestData;
};
const getDeepStackObject = (prediction) => {
if(prediction === undefined) {
return null;
}
const tag = prediction[detectorSettings.key];
const confidence = prediction.confidence ?? 0;
const y_min = prediction.y_min ?? 0;
const x_min = prediction.x_min ?? 0;
const y_max = prediction.y_max ?? 0;
const x_max = prediction.x_max ?? 0;
const width = x_max - x_min;
const height = y_max - y_min;
const obj = {
x: x_min,
y: y_min,
width: width,
height: height,
tag: tag,
confidence: confidence
};
if (detectorSettings.type === DETECTOR_TYPE_FACE) {
const matchingPersons = detectorSettings.registeredPersons.filter(p => tag.startsWith(p))
const person = matchingPersons.length > 0 ? matchingPersons[0] : null;
obj["person"] = person;
}
return obj;
};
initialize();
|
2022-12-11 08:53:04 +02:00 |
Moe
Antonín Skala
Katie Roberts