image_vision

Description:

image vision

Image Vision #

Offline image tagging & Face recognition #


+5000 Labels


Offline


Pre-train AI Model


Support PNG, JPEG, JPG and GIF


Powered by TFLite & Google ML Kit


Why was this plugin written?
The strength of this plugin is tagging photos on the device without communicating with anything outside the user's device. Well, the conclusion is that the most important issue for all of us is privacy. Initially, this plugin started with this vision, but it focuses on three main themes. Speed, accuracy and privacy security
What you need ?
In this example you need add Imager picker plugin
How to use Vision?
Simply

// image_picker plugin
import "dart:developer" as dev;
final ImagePicker picker = ImagePicker();

// Initial vision
var initialed = false ;
Future<bool> init() async {
initialed = await ImageVision.initial();
}
// Tag image function
Future<List<Map<String, dynamic>>> getLabels(File file) async {
var bytes = await file.readAsBytes();
String jsonLabels = await ImageVision.getTagsOfImage(Uint8List.fromList(bytes.toList()), 0.3);
var labels = List<Map<String, dynamic>>.from(json.decode(jsonLabels));
if (kDebugMode){
dev.log(labels.toString());
}
return labels ;
}

// Detect faces position on image
Future<List<Map<String, dynamic>>> getFaces(File file) async {
var bytes = await file.readAsBytes();
String jsonLabels = await ImageVision.detectFacesFromImage(Uint8List.fromList(bytes.toList()));
var faces = List<Map<String, dynamic>>.from(json.decode(jsonLabels));
if (kDebugMode){
dev.log(faces.toString());
}
return faces ;
}

// crop & recogniton face image
Future<dynamic> recognizeFace(Map<String, dynamic> inputFace, File image) async {
var face = img.decodeImage(await image.readAsBytes());
if (face != null){
face = img.copyCrop(
face,
int.parse(inputFace["left"].toString()),
int.parse(inputFace["top"].toString()),
int.parse(inputFace["width"].toString()),
int.parse(inputFace["height"].toString()),
);
final png = img.encodePng(face);
var rec = await ImageVision.recognizeFace(Uint8List.fromList(png));
var split = rec["confidence"].toString().split(".");
var number = split[0];
if (int.parse(number) < 1){
rec["title"] = "face_not_found";
rec["confidence"] = "0.0";
return rec;
} else {
return rec ;
}
}

return {};
}

// Register new face in Mobile Face net.
Future<String> register(String name, Map<String, dynamic> inputFace, File image) async {

var face = img.decodeImage(await image.readAsBytes());
if (face != null){
face = img.copyCrop(
face,
int.parse(inputFace["left"].toString()),
int.parse(inputFace["top"].toString()),
int.parse(inputFace["width"].toString()),
int.parse(inputFace["height"].toString()),
);
final png = img.encodePng(face);
var rec = await ImageVision.registerFace(name, Uint8List.fromList(png));
dev.log(rec.toString());
return rec ;
}
return "error";
}

copied to clipboard
Example response of getTagsOfImage() Method ...
[
{
"index":567,
"confidence":0.7956880331039429,
"label":"/m/01gq53",
"description":"Performance"

},
{
"index":2629,
"confidence":0.7695709466934204,
"label":"/m/04_5hy",
"description":"Stage"
},
{
"index":2990,
"confidence":0.32111555337905884,
"label":"/m/0557q",
"description":"Musical theatre"
}
]
copied to clipboard
Example response of detectFacesFromImage() Method ...
[
{
"top": 799,
"left": 686,
"right": 848,
"bottom": 961,
"width": 162,
"height": 162
},
{
"top": 801,
"left": 189,
"right": 416,
"bottom": 1024,
"width": 227,
"height": 223
},
{
"top": 840,
"left": 408,
"right": 536,
"bottom": 968,
"width": 128,
"height": 128
}
]
copied to clipboard
Enjoy :) #
Created with ❤️🍰☕ at Sensifai
Smart gallery uses Image Vision