# Object Detection using node-red-contrib-tfjs-coco-ssd

**URL:** <https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931>\
**Category:** General\
**Created:** [10 March 2020 14:41 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931 "2020-03-10T14:41:53Z")\
**Posts on this page:** 20\
**Page:** 6

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [12 April 2020 04:25 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/102 "2020-04-12T04:25:16Z")

</div>

Required nodes in addition to standard nodes (I use the latest current versions):

- node-red-contrib-image-tools
- node-red-contrib-telegrambot
- node-red-contrib-tfjs-coco-ssd
- node-red-dashboard
- node-red-node-base64

```auto
[{"id":"70f9130a.1f5c0c","type":"base64","z":"2d489fd9.1eedd","name":"","action":"str","property":"image","x":880,"y":530,"wires":[["af833380.f5ea9"]]},{"id":"420b8a89.423b94","type":"http request","z":"2d489fd9.1eedd","name":"","method":"GET","ret":"bin","paytoqs":false,"url":"https://loremflickr.com/320/240/sport","tls":"","persist":false,"proxy":"","authType":"","x":410,"y":430,"wires":[["62e85150.02f7b"]]},{"id":"77fd828a.a6073c","type":"ui_button","z":"2d489fd9.1eedd","name":"","group":"92518872.4d2bc8","order":3,"width":2,"height":1,"passthru":true,"label":"New Picture","tooltip":"","color":"","bgcolor":"","icon":"","payload":"","payloadType":"str","topic":"","x":260,"y":430,"wires":[["420b8a89.423b94"]]},{"id":"18eefa5b.18e276","type":"comment","z":"2d489fd9.1eedd","name":"========== ========== TFJS COCO SSD ========== ==========","info":"\n ","x":340,"y":490,"wires":[]},{"id":"7c8df2eb.c7408c","type":"inject","z":"2d489fd9.1eedd","name":"test","topic":"","payload":"","payloadType":"date","repeat":"","crontab":"","once":false,"onceDelay":0.1,"x":130,"y":430,"wires":[["77fd828a.a6073c"]]},{"id":"d0f460b0.4bdfd","type":"change","z":"2d489fd9.1eedd","name":"pay to detect","rules":[{"t":"move","p":"payload","pt":"msg","to":"detect","tot":"msg"}],"action":"","property":"","from":"","to":"","reg":false,"x":440,"y":530,"wires":[["d6827244.41d64","7a793272.3245fc"]]},{"id":"506351f8.e4c8","type":"link in","z":"2d489fd9.1eedd","name":"coco","links":["23ae07fd.88e498","c710c20f.8d8cb","ed0b6b69.870e78","3d40617c.70e06e"],"x":195,"y":530,"wires":[["e75695c9.dca4e8"]]},{"id":"3d40617c.70e06e","type":"link out","z":"2d489fd9.1eedd","name":"","links":["506351f8.e4c8"],"x":715,"y":430,"wires":[]},{"id":"3ef708a8.692668","type":"comment","z":"2d489fd9.1eedd","name":"coco","info":"","x":130,"y":530,"wires":[]},{"id":"e75695c9.dca4e8","type":"tensorflowCoco","z":"2d489fd9.1eedd","name":"","scoreThreshold":"","passthru":true,"x":290,"y":530,"wires":[["d0f460b0.4bdfd"]]},{"id":"d6827244.41d64","type":"function","z":"2d489fd9.1eedd","name":"Map Objects Boxes get.Picture","func":"//========== retrieves and transmits the image ==========\n//var pictureBuffer = flow.get('pictureBuffer')|| [];\n//msg.pictureBuffer = pictureBuffer; // send the picture\n\n//========== Detect Empty Objects ==========\nif(empty(msg.classes)){\nmsg.payload = \"no object detected\" //prepare for the google translate\nmsg.class=\"no object detected\" //prepare for display in Template Dashboard\nmsg.detect=[] //\nreturn msg\n} //no found object \nelse {\n// prepare the color and the thickness of the line in fct of the image size\nmsg.boxcolor = \"yellow\";\nmsg.textcolor = \"yellow\";\n \nif (msg.shape[1] < 300) {msg.textfontsize =\"10px\";msg.boxstroke =1;msg.textstroke = \"3px\";}\nelse if (msg.shape[1] >= 300 && msg.shape[1] < 500) {msg.textfontsize =\"15px\";msg.boxstroke =2;msg.textstroke = \"3px\";}\nelse if (msg.shape[1] >= 500 && msg.shape[1] < 900) {msg.textfontsize =\"20px\";msg.boxstroke =2;msg.textstroke = \"5px\";}\nelse if (msg.shape[1] >= 900 && msg.shape[1] < 2000){msg.textfontsize =\"50px\";msg.boxstroke =5;msg.textstroke = \"10px\";}\nelse if (msg.shape[1] >= 2000) {msg.textfontsize =\"80px\";msg.boxstroke =10;msg.textstroke = \"20px\";}\n\n//msg.textfontsize =(msg.shape[1] > 600) ? \"70px\" : \"10px\";\n//msg.boxstroke = (msg.shape[1] > 600) ? 5 : 2;\n//msg.textstroke = (msg.shape[1] > 600) ? \"10px\" : \"2px\";\n\n//========== indicates the type of objects and the quantity ==========\n// Get the array of names\nvar names = Object.keys(msg.classes);\nvar firstCount;\n\n// For each name, map it to \"n name\"\nvar parts = names.map((n,i) => {\n var count = msg.classes[n];\n if (i === 0) {\n //Remember the first count to get the \"is/are\" right later\n firstCount = count;\n }\n // Return \"n name\" and get the pluralisation right\n return count+\" \"+n+(count>1?\"s\":\"\")\n})\n// If there is more than one name, pop off the last one for later\nvar lastName;\nif (parts.length > 1) {\n lastName = parts.pop();\n}\n// Build up the response getting \"is/are\" right for the first count and joining\n// the array of names with a comma\nmsg.payload = \"There \"+(firstCount === 1 ? \"is\":\"are\")+\" \"+parts.join(\", \")\n// If there was a last name, add that on the end with an 'and', not a comma\nif (lastName) {\n msg.payload += \" and \"+lastName;\n}\nreturn msg;\n}//fin de si detecte\n\n//========== EMPTY FONCTION ==========\n/*\nHere's a simpler(short) solution to check for empty variables. \nThis function checks if a variable is empty. \nThe variable provided may contain mixed values (null, undefined, array, object, string, integer, function).\n*/\nfunction empty(mixed_var) {\n if (!mixed_var || mixed_var == '0') { return true; }\n if (typeof mixed_var == 'object') { \n for (var k in mixed_var) { return false; }\n return true; \n }\nreturn false;\n}//EMPTY FONCTION END\n","outputs":1,"noerr":0,"x":680,"y":530,"wires":[["70f9130a.1f5c0c"]]},{"id":"62e85150.02f7b","type":"change","z":"2d489fd9.1eedd","name":"","rules":[{"t":"set","p":"topic","pt":"msg","to":"testDetection","tot":"str"}],"action":"","property":"","from":"","to":"","reg":false,"x":570,"y":430,"wires":[["3d40617c.70e06e"]]},{"id":"c241e6ec.94fda8","type":"comment","z":"2d489fd9.1eedd","name":"coco","info":"","x":780,"y":430,"wires":[]},{"id":"7defeef9.85baf","type":"comment","z":"2d489fd9.1eedd","name":"========== UPLOAD picture ==========","info":"new picture :320x240\n\n0: object\n bbox: array[4]\n 0: 127\n 1: 86\n 2: 29\n 3: 90\n class: \"person\"\n score: 0.9015087080001831\n \n x=127 y=86 width =29 height=90\n ","x":250,"y":380,"wires":[]},{"id":"af833380.f5ea9","type":"ui_template","z":"2d489fd9.1eedd","group":"92518872.4d2bc8","name":"D3 templatev3 TextStroke","order":1,"width":7,"height":5,"format":"<head>\n <style>\n :root {\n --boxcolor: {{msg.boxcolor}};\n --boxstroke: {{msg.boxstroke}};\n --textcolor: {{msg.textcolor}};\n --textfontsize: {{msg.textfontsize}};\n --textstroke: {{msg.textstroke}};\n }\n .imag {\n width: 100%;\n height: 100%;\n } \n #svgimage text {\n font-family: Arial;\n font-size: var(--textfontsize, 18px);\n fill : var(--textcolor, yellow);\n paint-order: stroke;\n stroke: black;/*#ffffff;*//*yellow;*/\n stroke-width: var(--textstroke, 3px);/*3px;*/\n font-weight: 600;\n }\n rect {\n fill: blue;\n fill-opacity: 0;\n stroke: var(--boxcolor, yellow);\n stroke-width: var(--boxstroke, 1);\n }\n #svgimage {\n /*background-color: #cccccc; Used if the image is unavailable */\n background-repeat: no-repeat;\n background-size: cover;\n }\n </style>\n</head>\n<body>\n <svg preserveAspectRatio=\"xMidYMid meet\" id=\"svgimage\" style=\"width:100%\" viewBox=\"0 0 {{msg.shape[1]}} {{msg.shape[0]}}\">\n <image class=\"imag\" href=\"data:image/jpg;base64,{{msg.image}}\"/>\n </svg>\n <!-- \n <svg preserveAspectRatio=\"xMidYMid meet\" id=\"svgimage\" style=\"width:100%\" viewBox=\"0 0 {{msg.shape[1]}} {{msg.shape[0]}}\">\n <image class=\"imag\" href=\"data:image/jpg;base64,{{msg.pictureBuffer}}\"/>\n </svg> \n \n -->\n <div>{{msg.payload}}</div>\n <script>\n (function (scope) {\n scope.$watch('msg', function (msg) {\n if (msg && msg.detect) {\n var svg = d3.select(\"#svgimage\");\n\n var box = svg.selectAll(\"rect\").data(msg.detect)\n .attr(\"x\", function (d) { return d.bbox[0]; })\n .attr(\"y\", function (d) { return d.bbox[1]; })\n .attr(\"width\", function (d) { return d.bbox[2]; })\n .attr(\"height\", function (d) { return d.bbox[3]; });\n box.enter()\n .append(\"rect\")\n .attr(\"x\", function (d) { return d.bbox[0]; })\n .attr(\"y\", function (d) { return d.bbox[1]; })\n .attr(\"width\", function (d) { return d.bbox[2]; })\n .attr(\"height\", function (d) { return d.bbox[3]; });\n box.exit().remove();\n \n var text = svg.selectAll(\"text\").data(msg.detect)\n .text(function (d) { return d.class; })\n .attr(\"x\", function (d) { return d.bbox[0]; })\n .attr(\"y\", function (d) { return 10 + d.bbox[1]; });\n text.enter()\n .append(\"text\")\n .text(function (d) { return d.class; })\n .attr(\"x\", function (d) { return d.bbox[0]; })\n .attr(\"y\", function (d) { return 10 + d.bbox[1]; });\n text.exit().remove();\n }\n });\n })(scope);\n </script>\n</body>","storeOutMessages":true,"fwdInMessages":true,"templateScope":"local","x":1060,"y":530,"wires":[["6620cc3f.200064"]]},{"id":"7a793272.3245fc","type":"link out","z":"2d489fd9.1eedd","name":"","links":["892922d6.71f3a","d4e37092.e9a51"],"x":565,"y":580,"wires":[]},{"id":"86e7511.53999b","type":"jimp-image","z":"2d489fd9.1eedd","name":"","data":"image","dataType":"msg","ret":"img","parameter1":"","parameter1Type":"msg","parameter2":"","parameter2Type":"msg","parameter3":"","parameter3Type":"msg","parameter4":"","parameter4Type":"msg","parameter5":"","parameter5Type":"msg","parameter6":"","parameter6Type":"msg","parameter7":"","parameter7Type":"msg","parameter8":"","parameter8Type":"msg","parameterCount":0,"jimpFunction":"none","selectedJimpFunction":{"name":"none","fn":"none","description":"Just loads the image.","parameters":[]},"x":980,"y":660,"wires":[["3797d0ec.4ce87"]]},{"id":"bdfde2a1.a7e7d","type":"ui_template","z":"2d489fd9.1eedd","group":"7b6a751b.c5eb9c","name":"Canvas Selector","order":5,"width":1,"height":1,"format":" <script>\n (function(scope) {\n scope.$watch('msg', function(msg) {\n if (msg) {\n // Do something when msg arrives\n let dataURL = document.querySelector(\"#canvasImage\").toDataURL('image/jpeg');\n scope.send({payload: dataURL, w: msg.w, h: msg.h});\n }\n });\n })(scope);\n </script>\n","storeOutMessages":false,"fwdInMessages":false,"templateScope":"local","x":290,"y":780,"wires":[["7643a96c.565538"]]},{"id":"7643a96c.565538","type":"switch","z":"2d489fd9.1eedd","name":"data:image","property":"payload","propertyType":"msg","rules":[{"t":"cont","v":"data:image","vt":"str"}],"checkall":"true","repair":false,"outputs":1,"x":290,"y":840,"wires":[["e60214af.6cc108"]]},{"id":"d4e37092.e9a51","type":"link in","z":"2d489fd9.1eedd","name":"","links":["7a793272.3245fc"],"x":115,"y":660,"wires":[["97bf55b7.ec4198"]]},{"id":"c06dcfae.24d1b","type":"ui_template","z":"2d489fd9.1eedd","group":"7b6a751b.c5eb9c","name":"Canvas Template","order":7,"width":1,"height":1,"format":"<body>\n<canvas id=\"canvasImage\" width=\"1280\" height=\"720\">\n Your browser does not support the canvas element.\n</canvas>\n\n<script>\n (function(scope) {\n scope.$watch('msg', function (msg) {\n if (msg) {\n // Do something when msg arrives\n function squares(canvasid, squarelist, myimag) {\n let canvas = document.querySelector(canvasid);\n // hide the canvas\n canvas.style.display=\"none\";\n canvas.width = msg.shape[1];\n canvas.height = msg.shape[0];\n let ctx = canvas.getContext('2d');\n ctx.scale(1, 1);\n let image = new Image;\n image.src = \"data:image/jpg;base64,\"+myimag;\n image.onload = function () {\n ctx.drawImage(image, 0, 0, msg.shape[1], msg.shape[0]);\n for (const square of squarelist) {\n ctx.lineWidth = 2;\n ctx.strokeStyle = 'yellow';\n ctx.strokeRect(square.bbox[0], square.bbox[1], square.bbox[2], square.bbox[3]);\n ctx.font = \"10px Arial\";\n ctx.fillStyle = 'yellow';\n let width = ctx.measureText(square.class).width;\n ctx.fillRect(square.bbox[0], square.bbox[1]+square.bbox[3], square.bbox[2], 12);\n ctx.fillStyle = 'red';\n ctx.fillText(square.class, square.bbox[0]+square.bbox[2]/2-width/2, square.bbox[1]+square.bbox[3]+8);\n }\n };\n }\n\n let canvasid1 = \"#canvasImage\";\n let detect = msg.detect;\n let imag1 = msg.image;\n scope.send({payload: detect});\n squares(canvasid1, detect, imag1);\n scope.send({payload: 'go', w: msg.shape[1], h: msg.shape[0]});\n }\n });\n})(scope);\n</script>\n\n</body>","storeOutMessages":false,"fwdInMessages":false,"templateScope":"local","x":290,"y":720,"wires":[["bdfde2a1.a7e7d"]]},{"id":"97bf55b7.ec4198","type":"base64","z":"2d489fd9.1eedd","name":"","action":"str","property":"image","x":290,"y":660,"wires":[["c06dcfae.24d1b","c445e6a1.625ed8"]]},{"id":"892922d6.71f3a","type":"link in","z":"2d489fd9.1eedd","name":"","links":["7a793272.3245fc"],"x":825,"y":660,"wires":[["86e7511.53999b"]]},{"id":"b5d5d39e.401d9","type":"mqtt in","z":"2d489fd9.1eedd","name":"","topic":"imagetf","qos":"2","datatype":"auto","broker":"2a019090.5ba4d","x":490,"y":380,"wires":[["2dd4729f.4e239e"]]},{"id":"2dd4729f.4e239e","type":"switch","z":"2d489fd9.1eedd","name":"","property":"payload","propertyType":"msg","rules":[{"t":"istype","v":"buffer","vt":"buffer"}],"checkall":"true","repair":false,"outputs":1,"x":600,"y":380,"wires":[["3d40617c.70e06e"]]},{"id":"e60214af.6cc108","type":"ui_template","z":"2d489fd9.1eedd","group":"7b6a751b.c5eb9c","name":"SVG Template","order":2,"width":12,"height":9,"format":"<head>\n<style>\n .imag {\n width: 100%;\n height: 100%;\n } \n</style>\n</head>\n<body>\n <svg preserveAspectRatio=\"xMidYMid meet\" id=\"svgimage\" style=\"width:100%\" viewBox=\"0 0 {{msg.w}} {{msg.h}}\">\n <image class=\"imag\" href=\"{{msg.payload}}\"/>\n </svg>\n<script>\n(function(scope) {\n scope.$watch('msg', function(msg) {\n if (msg) {\n // Do something when msg arrives\n //scope.send({payload: dataURL});\n //alert(msg.w);\n }\n });\n})(scope);\n</script>\n</body>\n","storeOutMessages":true,"fwdInMessages":true,"templateScope":"local","x":290,"y":900,"wires":[["7aac40f.94f5bc"]]},{"id":"3797d0ec.4ce87","type":"function","z":"2d489fd9.1eedd","name":"draw Rects","func":"const LINE_WIDTH = 2;\nconst LINE_COLOR = 0xED143DFF;\n\nvar jimpImage = msg.payload;\nlet imgW = jimpImage.bitmap.width;\nlet imgH = jimpImage.bitmap.height;\n\n//create clamping functions to avoid printing outside of image\nlet clampX = (val) => clamp(val,0,imgW);\nlet clampY = (val) => clamp(val,0,imgH);\n\n//add imgBatchOps to msg for drawing text in next node \nmsg.imgBatchOps = [];\n\n\nif (msg.detect.length>0) {\n drawBoxes(jimpImage, msg.detect);\n}\n\nfunction drawBox(img, box) {\n //get box ccords\n let x = box.bbox[0];\n let y = box.bbox[1];\n let w = box.bbox[2];\n let h = box.bbox[3];\n \n scanHLine(img, x, y+h-14, w, 14, 0xfa758e22); //draw box for text\n scanRectangle(img, x, y, w, h, LINE_WIDTH, LINE_COLOR);//draw outer rect\n \n //built batch operations for next image node to draw text\n msg.imgBatchOps.push( \n {\n \"name\": \"print\",\n \"parameters\": [\n \"FONT_SANS_10_BLACK\",\n clampX(x+2),\n clampY(y+h-16),\n box.class\n]\n }\n \n ); \n}\n\nfunction drawBoxes(img, detections) {\n detections.forEach(element => drawBox(img, element))\n}\n\n\nfunction makeColorIterator(color) {\n return function (x, y, offset) {\n this.bitmap.data.writeUInt32BE(color, offset, true);\n }\n}\n\nfunction scanRectangle(image,x,y,w,h,linePX,color){\n let iterator = makeColorIterator(color);\n \n let xw, yh;\n x = clampX(x-(linePX/2));\n y = clampY(y-(linePX/2));\n w = clampX(w);\n h = clampY(h);\n xw = clampX(x+w);\n yh = clampY(y+h);\n\n \n if(y+linePX <= imgH){\n image.scan(x, y, w, linePX, iterator);// scan linePX height line - TOP\n }\n if(yh+linePX <= imgH){\n image.scan(x, yh, w, linePX, iterator);// scan linePX height line - BOTTOM\n }\n if(x+linePX <= imgW){\n image.scan(x, y, linePX, h, iterator);// scan linePX width line - LEFT\n }\n if(xw+linePX <= imgW){\n image.scan(xw, y, linePX, h, iterator);// scan linePX width line - RIGHT\n }\n \n}\n\n\nfunction scanHLine(image,x,y,length,linePX,color){\n let iterator = makeColorIterator(color);\n x = clampX(x);\n y = clampY(y);\n length = clampX(length);\n if(y+linePX <= imgH){\n image.scan(x, y, length, linePX, iterator);// \n }\n}\n\nfunction scanVLine(image,x,y,height,linePX,color){\n let iterator = makeColorIterator(color);\n x = clampX(x);\n y = clampY(y);\n height = clampY(y+height);\n if(x+linePX <= imgW){\n image.scan(x, y, linePX, height, iterator);// \n }\n}\n\n\n//scanCircle(jimpImage, 92, 170, 43, 0x000000FF);\n//scanCircle(jimpImage, 92, 170, 42, 0xFFCC00FF);\n\nfunction clamp(value, min, max) {\n return Math.min(Math.max(value, min), max);\n}\n\nfunction scanCircle(image, x, y, radius, color) {\n let iterator = makeColorIterator(color);\n return image.scan(x - radius, y - radius, radius*2, radius*2, function (pixX, pixY, idx) {\n if (Math.pow(pixX - x, 2) + Math.pow(pixY - y, 2) < radius*radius) {\n iterator.call(this, pixX, pixY, idx);\n }\n });\n}\n\n \n\nreturn msg;","outputs":1,"noerr":0,"x":980,"y":720,"wires":[["caf53b8c.5f6138"]]},{"id":"caf53b8c.5f6138","type":"jimp-image","z":"2d489fd9.1eedd","name":"add text","data":"payload","dataType":"msg","ret":"img","parameter1":"imgBatchOps","parameter1Type":"msg","parameter2":"","parameter2Type":"msg","parameter3":"","parameter3Type":"msg","parameter4":"","parameter4Type":"msg","parameter5":"","parameter5Type":"msg","parameter6":"","parameter6Type":"msg","parameter7":"","parameter7Type":"msg","parameter8":"","parameter8Type":"msg","parameterCount":1,"jimpFunction":"batch","selectedJimpFunction":{"name":"batch","fn":"batch","description":"apply one or more functions","parameters":[{"name":"options","type":"json","required":true,"hint":"an object or an array of objects containing {\"name\" : \"function_name\", \"parameters\" : [x,y,z]}. Refer to info on side panel}"}]},"x":980,"y":780,"wires":[["fd02639.baffca","526acc16.b9dfe4"]]},{"id":"5da5b688.b89a88","type":"telegram sender","z":"2d489fd9.1eedd","name":"","bot":"7e82026b.614f0c","x":980,"y":1200,"wires":[[]]},{"id":"c4973d1a.24498","type":"function","z":"2d489fd9.1eedd","name":"","func":"if ('person' in msg.classes){\n var m = { content:'', type:'', chatId:''};\n m.content = msg.payload;\n m.type = 'photo';\n m.chatId = 1234567890;\n msg.payload = m;\n return msg; \n}\n\n","outputs":1,"noerr":0,"x":980,"y":1140,"wires":[["5da5b688.b89a88"]]},{"id":"da054679.713928","type":"base64","z":"2d489fd9.1eedd","name":"","action":"","property":"payload","x":290,"y":1020,"wires":[[]]},{"id":"fd02639.baffca","type":"jimp-image","z":"2d489fd9.1eedd","name":"","data":"payload","dataType":"msg","ret":"b64","parameter1":"","parameter1Type":"msg","parameter2":"","parameter2Type":"msg","parameter3":"","parameter3Type":"msg","parameter4":"","parameter4Type":"msg","parameter5":"","parameter5Type":"msg","parameter6":"","parameter6Type":"msg","parameter7":"","parameter7Type":"msg","parameter8":"","parameter8Type":"msg","parameterCount":0,"jimpFunction":"none","selectedJimpFunction":{"name":"none","fn":"none","description":"Just loads the image.","parameters":[]},"x":980,"y":840,"wires":[["21c0651b.0bc42a"]]},{"id":"21c0651b.0bc42a","type":"ui_template","z":"2d489fd9.1eedd","group":"7b6a751b.c5eb9c","name":"SVG Template","order":1,"width":12,"height":9,"format":"<head>\n <style>\n .imag {\n width: 100%;\n height: 100%;\n } \n </style>\n</head>\n<body>\n <svg preserveAspectRatio=\"xMidYMid meet\" id=\"svgimage\" style=\"width:100%\" viewBox=\"0 0 {{msg.shape[1]}} {{msg.shape[0]}}\">\n <image class=\"imag\" href=\"{{msg.payload}}\"/>\n </svg>\n</body>","storeOutMessages":true,"fwdInMessages":true,"templateScope":"local","x":980,"y":900,"wires":[["81c1a54d.5c19a8"]]},{"id":"4b8734ac.3cb2ac","type":"base64","z":"2d489fd9.1eedd","name":"","action":"","property":"payload","x":980,"y":1080,"wires":[["c4973d1a.24498"]]},{"id":"81c1a54d.5c19a8","type":"split","z":"2d489fd9.1eedd","name":"","splt":",","spltType":"str","arraySplt":1,"arraySpltType":"len","stream":false,"addname":"payload","x":980,"y":960,"wires":[["d309fc02.ddea1"]]},{"id":"d309fc02.ddea1","type":"switch","z":"2d489fd9.1eedd","name":"","property":"payload","propertyType":"msg","rules":[{"t":"neq","v":"data:image","vt":"str"}],"checkall":"true","repair":false,"outputs":1,"x":980,"y":1020,"wires":[["4b8734ac.3cb2ac"]]},{"id":"1fd51178.32a59f","type":"image viewer","z":"2d489fd9.1eedd","name":"","width":"160","data":"payload","dataType":"msg","x":510,"y":960,"wires":[[]]},{"id":"526acc16.b9dfe4","type":"image viewer","z":"2d489fd9.1eedd","name":"","width":160,"data":"payload","dataType":"msg","x":1200,"y":780,"wires":[[]]},{"id":"c445e6a1.625ed8","type":"image viewer","z":"2d489fd9.1eedd","name":"","width":160,"data":"image","dataType":"msg","x":510,"y":660,"wires":[[]]},{"id":"6620cc3f.200064","type":"image viewer","z":"2d489fd9.1eedd","name":"","width":160,"data":"image","dataType":"msg","x":1230,"y":530,"wires":[[]]},{"id":"7aac40f.94f5bc","type":"function","z":"2d489fd9.1eedd","name":"","func":"msg.payload = msg.payload.split(',')[1];\nreturn msg;","outputs":1,"noerr":0,"x":290,"y":960,"wires":[["da054679.713928","1fd51178.32a59f"]]},{"id":"92518872.4d2bc8","type":"ui_group","z":"","name":"Video Capture","tab":"17522d42.149913","disp":false,"width":"7","collapse":false},{"id":"7b6a751b.c5eb9c","type":"ui_group","z":"","name":"NR Image tests","tab":"a0ac7d1b.a925d","disp":false,"width":24,"collapse":false},{"id":"2a019090.5ba4d","type":"mqtt-broker","z":"","name":"","broker":"192.168.0.240","port":"1883","clientid":"","usetls":false,"compatmode":true,"keepalive":"60","cleansession":true,"birthTopic":"","birthQos":"0","birthPayload":"","closeTopic":"","closePayload":"","willTopic":"","willQos":"0","willPayload":""},{"id":"7e82026b.614f0c","type":"telegram bot","z":"","botname":"Dummy","usernames":"","chatids":"","baseapiurl":"","updatemode":"polling","pollinterval":"300","usesocks":false,"sockshost":"","socksport":"6667","socksusername":"anonymous","sockspassword":"","bothost":"","localbotport":"8443","publicbotport":"8443","privatekey":"","certificate":"","useselfsignedcertificate":false,"sslterminated":false,"verboselogging":false},{"id":"17522d42.149913","type":"ui_tab","z":"","name":"Video Capture","icon":"dashboard","order":3,"disabled":false,"hidden":false},{"id":"a0ac7d1b.a925d","type":"ui_tab","z":"","name":"XXX","icon":"dashboard","order":1,"disabled":false,"hidden":false}]

```

As example, these images are handled excellent by the coco ssd analyze. Some other analyzers I have tested had problems detecting me walking there. The camera is looking down from the roof so no real frontal views that should have been ideal  
 ![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/c/4/c4106498b5b6c0bd6a30d3388ab3c2242d76d38f.jpeg)  
 ![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/0/4/04e33a465ce3dad99ff07ea53e7cbf0fc9f69db0.jpeg)

---

<div class="post-metadata">

**Author:** ![BartButenaers](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/bartbutenaers/32/10476_2.png) [@BartButenaers](https://discourse.nodered.org/u/BartButenaers)\
**Post date:** [12 April 2020 07:05 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/103 "2020-04-12T07:05:14Z")

</div>

Walter, this is awesome! And your flow is very compact.  
When I started with Node-RED a couple of years ago, this was the kind of stuff that I wanted to achieve.  
Finally we are getting closer to a pure Node-RED based video surveillance solution.  
Just love it 👍 👍 👍 👍 👍 👍

---

<div class="post-metadata">

**Author:** ![SuperNinja](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/superninja/32/5761_2.png) [@SuperNinja](https://discourse.nodered.org/u/SuperNinja)\
**Post date:** [12 April 2020 13:59 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/104 "2020-04-12T13:59:34Z")

</div>

a little bug that won't happen often. If we compare it to my flow :

 ![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/5/2/5257bf557eeb858b073ff34b534b5769a25d645a.jpeg)  
 ![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/1/f/1fd2fa725dfab5ee3f6efecda50e5a1e94a38325.png)

---

<div class="post-metadata">

**Author:** ![Steve-Mcl](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/steve-mcl/32/4826_2.png) [@Steve-Mcl](https://discourse.nodered.org/u/Steve-Mcl)\
**Post date:** [12 April 2020 14:04 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/105 "2020-04-12T14:04:40Z")

</div>

The problem is the very loose calculations for position of lines and rectangle for underneath text.

We don't have the luxury of a client side DOM with a plethora of elements and functions - at sever side, we have a bitmap of pixels and we have to calculate which pixels to set.

With a bit more work, we could easily determine form the rectangles coordinates in relation to the bitmap extent & place the string appropriately.

Edit...  
On the lighter side, we (humans) can see it's a person haha.

(the annotation is really only for visual debugging - I think 🙂 ).

---

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [12 April 2020 14:33 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/106 "2020-04-12T14:33:24Z")

</div>

I just have to tell you I have finally got something also working for my NVIDIA Jetson Nano's (arm64)!!! So happy!

I raised an issue earlier regarding missing support for node-tfjs on arm64 platforms

> <https://github.com/tensorflow/tfjs/issues/2872#event-3124096075>
>
> Hi, trying to install tensorflow.js to use with Node-RED in NVIDIA Jetson Nano. Node-RED itself runs perfect in the Nano
> Specifically I...

Luckily for me, @yhwang kindly reached out and provided a solution

> **[yhwang/node-red-contrib-tf-model](https://github.com/yhwang/node-red-contrib-tf-model#note)**
>
> A Node-RED node to load tensorflow model. Contribute to yhwang/node-red-contrib-tf-model development by creating an account on GitHub.

  
Maybe this is useful for RPi4 users as well?

Anyway, now I can run Node-RED on my Jetson Nano's with tfjs-coco-ssd detection built-in!

 ![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/3/5/35a2313e007550e0b2dbc094bd763b43535c9cf2.png)

---

<div class="post-metadata">

**Author:** ![Steve-Mcl](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/steve-mcl/32/4826_2.png) [@Steve-Mcl](https://discourse.nodered.org/u/Steve-Mcl)\
**Post date:** [12 April 2020 14:39 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/107 "2020-04-12T14:39:30Z")

</div>

Can you share this flow please?

---

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [12 April 2020 14:40 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/108 "2020-04-12T14:40:46Z")

</div>

Of course, right away...

---

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [12 April 2020 14:46 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/109 "2020-04-12T14:46:08Z")

</div>

I found out that some additional nodes also are needed. I installed them via palette manager before I imported the flow (image-tools you have already I think 😉 )

Best regards, Walter

"dependencies": {  
"@tensorflow/tfjs-node": "1.4.0",  
"node-red": "^1.0.3",  
"node-red-contrib-browser-utils": "0.0.9",  
"node-red-contrib-image-tools": "^0.2.5",  
"node-red-contrib-post-object-detection": "^0.1.2",  
"node-red-contrib-tf-function": "^0.1.0",  
"node-red-contrib-tf-model": "^0.1.6"

```auto
[{"id":"1545e935.999977","type":"jimp-image","z":"bfd6731.db3089","name":"load the image","data":"payload","dataType":"msg","ret":"buf","parameter1":"","parameter1Type":"msg","parameter2":"","parameter2Type":"msg","parameter3":"","parameter3Type":"msg","parameter4":"","parameter4Type":"msg","parameter5":"","parameter5Type":"msg","parameter6":"","parameter6Type":"msg","parameter7":"","parameter7Type":"msg","parameter8":"","parameter8Type":"msg","parameterCount":0,"jimpFunction":"none","selectedJimpFunction":{"name":"none","fn":"none","description":"Just loads the image.","parameters":[]},"x":360,"y":140,"wires":[["8721e9e0.e949e8","d775e36d.cc825","11e3829b.dec12d"]]},{"id":"8721e9e0.e949e8","type":"image viewer","z":"bfd6731.db3089","name":"Original Image viewer","width":"320","data":"payload","dataType":"msg","x":940,"y":140,"wires":[[]]},{"id":"5606b167.a042b","type":"image viewer","z":"bfd6731.db3089","name":"With bounding boxes","width":"320","data":"payload","dataType":"msg","x":660,"y":570,"wires":[[]]},{"id":"7ceb8aa5.98cbe4","type":"bbox-image","z":"bfd6731.db3089","name":"bounding-box","x":570,"y":500,"wires":[["5606b167.a042b"]]},{"id":"dd52b9b3.e03978","type":"change","z":"bfd6731.db3089","name":"objects","rules":[{"t":"set","p":"complete","pt":"msg","to":"true","tot":"bool"}],"action":"","property":"","from":"","to":"","reg":false,"x":360,"y":430,"wires":[["11e3829b.dec12d"]]},{"id":"bcf6963a.e2f4e8","type":"post-object-detection","z":"bfd6731.db3089","classesURL":"https://s3.sjc.us.cloud-object-storage.appdomain.cloud/tfjs-cos/cocossd/classes.json","iou":"0.5","minScore":"0.5","name":"post-processing","x":360,"y":360,"wires":[["dd52b9b3.e03978"]]},{"id":"d775e36d.cc825","type":"tf-function","z":"bfd6731.db3089","name":"pre-processing","func":"const image = tf.tidy(() => {\n return tf.node.decodeImage(msg.payload, 3).expandDims(0);\n});\n\nreturn {payload: { image_tensor: image } };","outputs":1,"noerr":0,"x":360,"y":210,"wires":[["8bee1dc1.4ecca"]]},{"id":"8bee1dc1.4ecca","type":"tf-model","z":"bfd6731.db3089","modelURL":"https://storage.googleapis.com/tfjs-models/savedmodel/ssdlite_mobilenet_v2/model.json","outputNode":"","name":"COCO SSD","x":350,"y":280,"wires":[["bcf6963a.e2f4e8"]]},{"id":"11e3829b.dec12d","type":"function","z":"bfd6731.db3089","name":"","func":"let queue = flow.get('queue');\nif (queue === undefined) {\n queue = [];\n flow.set('queue', queue);\n}\n\nif (msg.complete === undefined) {\n queue.push(msg.payload);\n node.done();\n} else {\n const image = queue.shift();\n node.send(\n {\n payload: {\n objects: msg.payload,\n image: image\n }\n });\n}","outputs":1,"noerr":0,"x":565,"y":430,"wires":[["7ceb8aa5.98cbe4"]],"icon":"node-red/join.svg","l":false},{"id":"f8135785.7c4808","type":"mqtt in","z":"bfd6731.db3089","name":"","topic":"epic","qos":"2","datatype":"auto","broker":"a5576de4.82b","x":140,"y":140,"wires":[["1545e935.999977"]]},{"id":"ff74da07.dd21f8","type":"fileinject","z":"bfd6731.db3089","name":"","x":150,"y":100,"wires":[["1545e935.999977"]]},{"id":"a5576de4.82b","type":"mqtt-broker","z":"","name":"","broker":"192.168.0.240","port":"1883","clientid":"","usetls":false,"compatmode":true,"keepalive":"60","cleansession":true,"birthTopic":"","birthQos":"0","birthPayload":"","closeTopic":"","closePayload":"","willTopic":"","willQos":"0","willPayload":""}]

```

---

<div class="post-metadata">

**Author:** ![Steve-Mcl](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/steve-mcl/32/4826_2.png) [@Steve-Mcl](https://discourse.nodered.org/u/Steve-Mcl)\
**Post date:** [12 April 2020 14:53 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/110 "2020-04-12T14:53:33Z")

</div>

Ah now I understand. It uses node-canvas for the drawing. So it inherits the canvas capabilities - good workaround for the limited drawing capabilities of jimp.

---

<div class="post-metadata">

**Author:** ![BartButenaers](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/bartbutenaers/32/10476_2.png) [@BartButenaers](https://discourse.nodered.org/u/BartButenaers)\
**Post date:** [12 April 2020 18:58 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/111 "2020-04-12T18:58:27Z")

</div>

> [@krambriw](#):
>
> Luckily for me, @yhwang kindly reached out and provided a solution

Do you mean you use now the tf-model node instead of the coco-ssd node? And is Jetson Nano a better choice (instead of a rpi 4) for this kind of stuff?

> [@krambriw](#):
>
> these images are handled excellent by the coco ssd analyze

The analysis is not always correct.  
E.g. the analysis of this [picture](http://1.bp.blogspot.com/-1Mcca9ySyco/UVdoLDMx9qI/AAAAAAAADmU/Yg2zlqJ9eWM/s1600/vw_kever_in_prak.PNG) results in a motorcycle instead of a car:

 ![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/0/0/005ba8a1b46ad0c8f9de7f37978c43aa60daf128.jpeg)

I have told my wife so many times that she should be careful when parking our Volkswagen 🤔

Sorry for this amateurish interruption of a professional discussion ...

---

<div class="post-metadata">

**Author:** ![SuperNinja](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/superninja/32/5761_2.png) [@SuperNinja](https://discourse.nodered.org/u/SuperNinja)\
**Post date:** [12 April 2020 19:38 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/112 "2020-04-12T19:38:32Z")

</div>

> [@BartButenaers](#):
>
> I have told my wife so many times that she should be careful when parking our Volkswagen 🤔

In addition, it was an almost new car! 😆

> [@BartButenaers](#):
>
> The analysis is not always correct.

Yes, I too see the same problem: plants are detected as People and create false alerts through the loudspeaker of the house at 2:00 am: "There is a person in the garden" 🥴  
**By doing research in the other examples of flow I found this one:**

> **[node-red-contrib-teachable-machine](https://flows.nodered.org/node/node-red-contrib-teachable-machine)**
>
> Simplifies integration with Teachable Machine models from Google

[https://miro.medium.com/max/1400/1\*h6Vz9fIQJRoASHOK0uKHGg.gif](https://miro.medium.com/max/1400/1*h6Vz9fIQJRoASHOK0uKHGg.gif)

It is based on Teachable Machin by google. Using a webcam we take the object from various angles, backgrounds, lighting ... To improve the model, we can add other photos.  
The result is a tensorflow.js model, the url of which must be entered in the detection node. Or registered locally. So simple.  
To test ...

---

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [13 April 2020 07:14 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/113 "2020-04-13T07:14:45Z")

</div>

> [@BartButenaers](#):
>
> Do you mean you use now the tf-model node instead of the coco-ssd node? And is Jetson Nano a better choice (instead of a rpi 4) for this kind of stuff?

Yes, currently, on the Jetson Nano I use the tf-model node and it is stated that the GPU is used. To make the coco-ssd-node work I suspect that it has to use the updated @tensorflow/tfs-node but I do not know how to make that happen, I have some ideas experimenting, like running "npm install" in node-red-contrib-tfjs-coco-ssd directory to rebuild the bindings, but I don't know

_EDIT: I do not know but I do suspect that it is problematic to use the node-red-contrib-tfjs-coco-ssd node on a RPi4. Maybe someone already tried?_

If the Jetson Nano is a better choice then the RPi4 for this kind? Well, I don't know. What I know is that the Nano can do impressive real time video analyzes at a high speed when using other tools & libraries specifically written for it. But I'm not sure if it's power is really utilized when running everything through NR.

> [@BartButenaers](#):
>
> careful when parking

So very true 😃  
On the other hand, my wife once worked with a company handling customer fleets (car leasing) and you cannot believe how many calls started "My wife...". In reality, it was not the wife...

---

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [13 April 2020 07:15 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/114 "2020-04-13T07:15:39Z")

</div>

> [@SuperNinja](#):
>
> To test ...

Very interesting!!!

---

<div class="post-metadata">

**Author:** ![SuperNinja](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/superninja/32/5761_2.png) [@SuperNinja](https://discourse.nodered.org/u/SuperNinja)\
**Post date:** [13 April 2020 07:51 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/115 "2020-04-13T07:51:27Z")

</div>

> [@krambriw](#):
>
> If the Jetson Nano is a better choice then the RPi4 for this kind?

 ![](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/f/f/ff2de957c0c1d5a7ba1c6b8ca22084f62ba0b487.png)  
_This decrease in inferencing time brings the Raspberry Pi 4 directly into competition with both the NVIDIA Jetson Nano and the Movidius-based hardware from Intel_

**For the most curious the link of the comparison (very instructive). The difference between Tensorflow and TensorflowLight is obvious:**

> **[Benchmarking TensorFlow Lite on the New Raspberry Pi 4, Model B](https://www.hackster.io/news/benchmarking-tensorflow-lite-on-the-new-raspberry-pi-4-model-b-3fd859d05b98)**
>
> When the Raspberry Pi 4 was launched I sat down to update the benchmarks I’ve been putting together for the new generation of accelerator…

---

<div class="post-metadata">

**Author:** ![krambriw](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/krambriw/32/5429_2.png) [@krambriw](https://discourse.nodered.org/u/krambriw)\
**Post date:** [13 April 2020 08:29 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/116 "2020-04-13T08:29:50Z")

</div>

> [@krambriw](#):
>
> I have some ideas experimenting

So I managed to make the coco-ssd-node work on Jetson Nano!!! This is how it worked for me:

Install both nodes:

- node-red-contrib-tf-model  
(following the guide referenced already above [GitHub - yhwang/node-red-contrib-tf-model: A Node-RED node to load tensorflow model](https://github.com/yhwang/node-red-contrib-tf-model#note) )
- node-red-contrib-tfjs-coco-ssd

In the folder "/home/user/.node-red/node\_modules/node-red-contrib-tfjs-coco-ssd/node\_modules/@tensorflow" delete the following folders

- tfjs-node
- tfjs-converter

and then replace with the same folders from the folder "/home/user/.node-red/node\_modules/@tensorflow"

After this the coco-ssd node also works on arm64!

![image](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/6/a/6adee7fe795c95263790c9ef3434c9ed844caa3f.png)

---

<div class="post-metadata">

**Author:** ![Jean-Luc](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/jean-luc/32/89996_2.png) [@Jean-Luc](https://discourse.nodered.org/u/Jean-Luc)\
**Post date:** [22 April 2020 16:19 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/117 "2020-04-22T16:19:02Z")

</div>

I was able to test the detection of objects but on the pi3 there is a memory consumption that increases without ever going down.  
After about 10 detections it blocks. It is necessary to restart NR.  
Was this already mentioned-resolved in another post? If so, sorry 🤭  
If not, do you have ideas to solve this problem?  
I attach a CPU & Memory visualization to each detection.

![mémoire-cpu](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/5/d/5d110788b39288b5b02ee6dea203fb2354a0687e.jpeg)

 ![terminal-log](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/5/0/50899aea59dd16af3d7f023550b61bde867dbdb1.jpeg)

![NR-LOG](https://us1.discourse-cdn.com/flex026/uploads/nodered/original/3X/4/7/4727084730815bcfd7841d4d8fa4e3dc831bffb5.jpeg)

---

<div class="post-metadata">

**Author:** ![wiredquill](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/wiredquill/32/19765_2.png) [@wiredquill](https://discourse.nodered.org/u/wiredquill)\
**Post date:** [22 April 2020 16:39 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/118 "2020-04-22T16:39:45Z")

</div>

I'm having the exact same issue in a RPi 4.

I had to write a flow that restarts Node Red once it it 60% or Ram consumed.

---

<div class="post-metadata">

**Author:** ![dceejay](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/dceejay/32/38_2.png) [@dceejay](https://discourse.nodered.org/u/dceejay)\
**Post date:** [22 April 2020 17:36 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/119 "2020-04-22T17:36:30Z")

</div>

Which version of the node are you using ? Can you try shrinking the images before sending them to the node. (While it shouldn’t matter the models only need about 320 x 240 px)

---

<div class="post-metadata">

**Author:** ![SuperNinja](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/superninja/32/5761_2.png) [@SuperNinja](https://discourse.nodered.org/u/SuperNinja)\
**Post date:** [22 April 2020 20:28 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/120 "2020-04-22T20:28:57Z")

</div>

here I have programmed a resizing of the images captured by the cameras at 320x240. Since 2 hours, I have not exceeded 90% of use (Ram 1 GB). I will let it run for a few days to see...

In an other hand, more and more people complain of memory leak when using tfjs:

> <https://github.com/tensorflow/tfjs/issues/1440>
>
> A while ago I filed the issue #604 with a description how variably sized input tensors lead to memory leaks utilizing...

  

> <https://stackoverflow.com/questions/56513495/memory-leak-in-tensorflow-js-how-to-clean-up-unused-tensors>

  
It seems that using the `td.tidy()` method and the `tf.disposeVariables` function solves the problem.  
@dceedjay does that mean anything to you? Can it be useful to solve our problem?

---

<div class="post-metadata">

**Author:** ![dceejay](https://sea2.discourse-cdn.com/flex026/user_avatar/discourse.nodered.org/dceejay/32/38_2.png) [@dceejay](https://discourse.nodered.org/u/dceejay)\
**Post date:** [22 April 2020 22:09 UTC](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931/121 "2020-04-22T22:09:34Z")

</div>

worth a try - pushed a version 0.4.1  
not sure if I've implemented it correctly - so any help / advice gratefully received...

[Previous page](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931.md?page=5)

[Next page](https://discourse.nodered.org/t/object-detection-using-node-red-contrib-tfjs-coco-ssd/22931.md?page=7)