224 行
17 KiB
Objective-C
224 行
17 KiB
Objective-C
#import <Foundation/Foundation.h>
|
|
#import <CoreImage/CoreImage.h>
|
|
#import <Metal/Metal.h>
|
|
#import "PGFaceProcessor.h"
|
|
#include <math.h>
|
|
|
|
@interface PGFaceProcessor (SlimTest)
|
|
- (CIImage *)warpImage:(CIImage *)image geometry:(NSArray<NSDictionary *> *)geometry settings:(PGBeautySettings)s extent:(CGRect)extent;
|
|
@end
|
|
@interface PGRealtimeFaceTracker : NSObject
|
|
- (NSDictionary *)resultForGeometry:(NSArray *)raw settings:(PGBeautySettings)s time:(CMTime)time extent:(CGRect)extent;
|
|
@end
|
|
static NSUInteger checks=0;
|
|
static double maxOracleError=0;
|
|
static void Check(BOOL ok,NSString *message) {checks++;if(!ok){fprintf(stderr,"FAIL: %s\n",message.UTF8String);exit(1);}}
|
|
static double Clamp(double x,double a,double b) {return fmin(b,fmax(a,x));}
|
|
static CGPoint ReadPoint(NSArray *a) {return CGPointMake([a[0] doubleValue],[a[1] doubleValue]);}
|
|
static NSArray *WritePoint(CGPoint p) {return @[@(p.x),@(p.y)];}
|
|
|
|
// Authored 2D projections, with the documented 76-point layout. They are not
|
|
// extracted vendor assets or private face coordinates. Positive yaw projects
|
|
// the nose toward the right while the silhouette becomes narrower.
|
|
static NSDictionary *Geometry(double yaw,double roll) {
|
|
NSMutableArray *p=[NSMutableArray new];
|
|
for(NSUInteger i=0;i<76;i++)[p addObject:@[@.5,@.5]];
|
|
p[6]=@[@.34,@.65];p[13]=@[@.66,@.65];
|
|
p[47]=@[@.5,@.60];p[48]=@[@.5,@.52];p[49]=@[@.5,@.45];p[52]=@[@.5,@.42];
|
|
p[36]=@[@.54,@.36];p[37]=@[@.5,@.35];p[38]=@[@.46,@.36];
|
|
for(NSUInteger i=0;i<17;i++) {
|
|
double t=M_PI+M_PI*i/16;
|
|
p[59+i]=@[@(.5+.29*cos(t)),@(.65+.43*sin(t))];
|
|
}
|
|
for(NSUInteger i=0;i<76;i++) {
|
|
CGPoint a=ReadPoint(p[i]);
|
|
double depth=(i==47 || i==48 || i==49 || i==52)?.18:.03;
|
|
double x=(a.x-.5)*cos(yaw)+depth*sin(yaw),y=a.y-.5;
|
|
p[i]=WritePoint(CGPointMake(.5+x*cos(roll)-y*sin(roll),.5+x*sin(roll)+y*cos(roll)));
|
|
}
|
|
NSMutableDictionary *g=[NSMutableDictionary dictionaryWithDictionary:@{@"box":@[@.15,@.15,@.7,@.6],@"points76":p}];
|
|
for(NSString *key in @[@"leftEye",@"rightEye",@"leftBrow",@"rightBrow",@"nose",@"lips",@"mouth",@"contour"])g[key]=@[];
|
|
return g;
|
|
}
|
|
|
|
// Independent scalar oracle from the recorded AIR specification. Its explicit
|
|
// y-down conversion checks the Core Image implementation's coordinate bridge.
|
|
static CGPoint Oracle(CGPoint pixel,NSDictionary *g,CGRect extent,double thin) {
|
|
static const unsigned int origins[17]={75,59,74,60,73,61,72,62,71,63,70,64,69,65,68,66,67};
|
|
static const unsigned int anchors[17]={6,13,47,47,47,47,48,48,48,48,49,49,49,49,52,52,47};
|
|
NSArray *raw=g[@"points76"];CGPoint p[76];
|
|
for(NSUInteger i=0;i<76;i++){p[i]=ReadPoint(raw[i]);p[i].y=1-p[i].y;}
|
|
double width=hypot((p[75].x-p[59].x)*extent.size.width,(p[75].y-p[59].y)*extent.size.height);
|
|
double height=hypot((p[67].x-(p[74].x+p[60].x)/2)*extent.size.width,(p[67].y-(p[74].y+p[60].y)/2)*extent.size.height);
|
|
double gain=Clamp(280/fmax(1,fmax(width,2*height)),1,2.25)*.025*Clamp(thin,0,1);
|
|
CGPoint q=CGPointMake((pixel.x-extent.origin.x)/extent.size.width,1-(pixel.y-extent.origin.y)/extent.size.height);
|
|
for(NSUInteger i=0;i<17;i++) {
|
|
CGPoint a=p[origins[i]],b=p[anchors[i]];
|
|
double r=hypot(a.x-b.x,a.y-b.y);
|
|
if(r<=.0001)continue;
|
|
double w=Clamp(1-hypot(q.x-a.x,q.y-a.y)/r,0,1);
|
|
q.x=Clamp(q.x-(b.x-a.x)*gain*w,0,1);q.y=Clamp(q.y-(b.y-a.y)*gain*w,0,1);
|
|
}
|
|
return CGPointMake(q.x,1-q.y);
|
|
}
|
|
static CGPoint ChinOracle(CGPoint pixel,NSDictionary *g,CGRect e,double thin,double chin) {
|
|
CGPoint q=Oracle(pixel,g,e,thin);q.y=1-q.y;
|
|
NSArray *raw=g[@"points76"];CGPoint points[76];
|
|
for(NSUInteger i=0;i<76;i++){points[i]=ReadPoint(raw[i]);points[i].y=1-points[i].y;}
|
|
double width=hypot((points[75].x-points[59].x)*e.size.width,(points[75].y-points[59].y)*e.size.height);
|
|
double height=2*hypot((points[67].x-(points[74].x+points[60].x)/2)*e.size.width,(points[67].y-(points[74].y+points[60].y)/2)*e.size.height);
|
|
double ratio=Clamp(chin,-1,1)*(chin>0?.04:.02)*Clamp(280/fmax(1,fmax(width,height)),1,2.25);
|
|
for(NSUInteger i=0;i<3;i++) {
|
|
CGPoint a=points[68-i],b=points[38-i];double r=hypot(a.x-b.x,a.y-b.y);
|
|
if(r<=.0001)continue;
|
|
double w=Clamp(1-hypot(q.x-a.x,q.y-a.y)/r,0,1);
|
|
q.x=Clamp(q.x-(b.x-a.x)*ratio*w,0,1);q.y=Clamp(q.y-(b.y-a.y)*ratio*w,0,1);
|
|
}
|
|
return CGPointMake(q.x,1-q.y);
|
|
}
|
|
static CIImage *CoordinateImage(CGRect extent) {
|
|
NSUInteger w=extent.size.width,h=extent.size.height;
|
|
NSMutableData *d=[NSMutableData dataWithLength:w*h*4*sizeof(float)];float *p=d.mutableBytes;
|
|
for(NSUInteger y=0;y<h;y++)for(NSUInteger x=0;x<w;x++) {
|
|
NSUInteger k=(y*w+x)*4;p[k]=(x+.5)/w;p[k+1]=(y+.5)/h;p[k+2]=.25;p[k+3]=1;
|
|
}
|
|
return [[CIImage imageWithBitmapData:d bytesPerRow:w*4*sizeof(float) size:extent.size format:kCIFormatRGBAf colorSpace:nil]
|
|
imageByApplyingTransform:CGAffineTransformMakeTranslation(extent.origin.x,extent.origin.y)];
|
|
}
|
|
static NSData *Pixels(CIContext *c,CIImage *image) {
|
|
CGRect e=image.extent;NSUInteger w=e.size.width,h=e.size.height;
|
|
NSMutableData *d=[NSMutableData dataWithLength:w*h*4*sizeof(float)];
|
|
[c render:image toBitmap:d.mutableBytes rowBytes:w*4*sizeof(float) bounds:e format:kCIFormatRGBAf colorSpace:nil];return d;
|
|
}
|
|
static NSData *Archive(NSDictionary *geometry) {
|
|
return [NSPropertyListSerialization dataWithPropertyList:@{@"version":@1,@"frames":@{@"0/1/0":@{@"value":@0,@"scale":@1,@"epoch":@0,@"faces":@[geometry]}}} format:NSPropertyListBinaryFormat_v1_0 options:0 error:nil];
|
|
}
|
|
static void NumericSettingsCase(CIContext *c,PGFaceProcessor *p,CGRect e,NSDictionary *g,double amount,double chin,NSString *name) {
|
|
PGBeautySettings s=PGBeautyDefaults();s.faceThin=amount;s.faceChin=chin;
|
|
CIImage *input=CoordinateImage(e),*output=[p warpImage:input geometry:@[g] settings:s extent:e];
|
|
Check(output!=nil && CGRectEqualToRect(output.extent,e),[name stringByAppendingString:@" preserves extent"]);
|
|
NSData *data=Pixels(c,output);const float *pixels=data.bytes;double worst=0,maxMove=0;
|
|
NSUInteger worstX=0,worstY=0;CGPoint worstExpected=CGPointZero,worstActual=CGPointZero;
|
|
NSUInteger w=e.size.width,h=e.size.height;
|
|
for(NSUInteger y=3;y<h-3;y+=7)for(NSUInteger x=3;x<w-3;x+=7) {
|
|
// Bitmap rows are top first, while destCoord / geometry are y up.
|
|
CGPoint q=ChinOracle(CGPointMake(e.origin.x+x+.5,e.origin.y+h-y-.5),g,e,amount,chin);
|
|
double ex=Clamp(q.x,.5/w,1-.5/w),ey=Clamp(1-q.y,.5/h,1-.5/h);
|
|
NSUInteger k=(y*w+x)*4;
|
|
double difference=fmax(fabs(pixels[k]-ex),fabs(pixels[k+1]-ey));
|
|
if(difference>worst){worst=difference;worstX=x;worstY=y;worstExpected=CGPointMake(ex,ey);worstActual=CGPointMake(pixels[k],pixels[k+1]);}
|
|
maxMove=fmax(maxMove,hypot(ex-(x+.5)/w,ey-(y+.5)/h));
|
|
}
|
|
if(worst>0.00006)fprintf(stderr,"%s oracle max error=%g at %lu,%lu expected=(%g,%g) actual=(%g,%g)\n",name.UTF8String,worst,(unsigned long)worstX,(unsigned long)worstY,worstExpected.x,worstExpected.y,worstActual.x,worstActual.y);
|
|
maxOracleError=fmax(maxOracleError,worst);
|
|
Check(worst<0.00006,[name stringByAppendingString:@" GPU field matches the independent thin/chin oracle"]);
|
|
Check((amount==0 && chin==0) || maxMove>.0001,[name stringByAppendingString:@" exercises a nonzero displacement"]);
|
|
NSUInteger corner=4*((h-3)*w+(w-3));
|
|
Check(fabs(pixels[corner]-(w-2.5)/w)<1e-6 && fabs(pixels[corner+1]-(h-2.5)/h)<1e-6,[name stringByAppendingString:@" preserves distant background"]);
|
|
}
|
|
static void NumericCase(CIContext *c,PGFaceProcessor *p,CGRect e,NSDictionary *g,double amount,NSString *name) {
|
|
NumericSettingsCase(c,p,e,g,amount,0,name);
|
|
}
|
|
static CGPoint Displacement(CIContext *c,CIImage *ramp,CIImage *warped,CGPoint uv) {
|
|
CGRect e=ramp.extent;CGRect point=CGRectMake(floor(e.origin.x+uv.x*e.size.width),floor(e.origin.y+uv.y*e.size.height),1,1);
|
|
float a[4],b[4];
|
|
[c render:ramp toBitmap:a rowBytes:sizeof(a) bounds:point format:kCIFormatRGBAf colorSpace:nil];
|
|
[c render:warped toBitmap:b rowBytes:sizeof(b) bounds:point format:kCIFormatRGBAf colorSpace:nil];
|
|
return CGPointMake((b[0]-a[0])*e.size.width,-(b[1]-a[1])*e.size.height);
|
|
}
|
|
static void AnatomicalChecks(CIContext *c,PGFaceProcessor *p,NSDictionary *g,CGRect e) {
|
|
CIImage *ramp=CoordinateImage(e);NSArray *points=g[@"points76"];
|
|
CGPoint chin=ReadPoint(points[67]),lip=ReadPoint(points[37]);
|
|
double ax=(lip.x-chin.x)*e.size.width,ay=(lip.y-chin.y)*e.size.height;
|
|
for(int sign=-1;sign<=1;sign+=2) {
|
|
PGBeautySettings s=PGBeautyDefaults();s.faceChin=sign;
|
|
CIImage *out=[p warpImage:ramp geometry:@[g] settings:s extent:e];
|
|
CGPoint d=Displacement(c,ramp,out,chin);
|
|
Check(hypot(d.x,d.y)>.05,@"chin control moves the detected physical chin landmark");
|
|
Check(sign*(d.x*ax+d.y*ay)<0,@"signed chin direction follows chin-to-lip anchor instead of a fixed image Y axis");
|
|
BOOL outside=YES,eyesUnchanged=YES;
|
|
for(NSNumber *index in @[@6,@13]) {
|
|
CGPoint eye=ReadPoint(points[index.unsignedIntegerValue]);
|
|
for(NSUInteger i=0;i<3;i++) {
|
|
CGPoint a=ReadPoint(points[68-i]),b=ReadPoint(points[38-i]);
|
|
if(hypot(eye.x-a.x,eye.y-a.y)<=hypot(a.x-b.x,a.y-b.y)+.003)outside=NO;
|
|
}
|
|
CGPoint shift=Displacement(c,ramp,out,eye);eyesUnchanged &= hypot(shift.x,shift.y)<.002;
|
|
}
|
|
Check(outside && eyesUnchanged,@"both physical pupils outside chin supports remain unchanged");
|
|
}
|
|
}
|
|
|
|
int main(int argc,const char *argv[]) {@autoreleasepool {
|
|
CIContext *c=[CIContext contextWithMTLDevice:MTLCreateSystemDefaultDevice() options:@{kCIContextWorkingColorSpace:NSNull.null,kCIContextOutputColorSpace:NSNull.null}];
|
|
PGFaceProcessor *p=[PGFaceProcessor new];
|
|
CGRect extents[]={CGRectMake(0,0,240,240),CGRectMake(0,0,320,180),CGRectMake(-17,29,180,320),CGRectMake(0,0,640,360)};
|
|
double yaw[]={0,.65,1.40,-1.40},roll[]={0,.45,-.50};
|
|
NumericCase(c,p,extents[1],Geometry(0,0),0,@"zero strength");
|
|
for(NSUInteger e=0;e<4;e++)for(NSUInteger y=0;y<4;y++)for(NSUInteger r=0;r<3;r++) {
|
|
NSString *name=[NSString stringWithFormat:@"size%lu yaw%lu roll%lu",(unsigned long)e,(unsigned long)y,(unsigned long)r];
|
|
NumericCase(c,p,extents[e],Geometry(yaw[y],roll[r]),1,name);
|
|
}
|
|
NumericCase(c,p,extents[1],Geometry(.6,.2),.5,@"half strength");
|
|
for(NSUInteger e=0;e<3;e++)for(NSUInteger y=0;y<3;y++)for(NSUInteger r=0;r<2;r++)for(int sign=-1;sign<=1;sign+=2) {
|
|
double angle=y==0?0:(y==1?1.4:-1.4);
|
|
NumericSettingsCase(c,p,extents[e],Geometry(angle,r?.45:0),0,sign,
|
|
[NSString stringWithFormat:@"chin size%lu yaw%lu roll%lu sign%d",(unsigned long)e,(unsigned long)y,(unsigned long)r,sign]);
|
|
}
|
|
NumericSettingsCase(c,p,extents[1],Geometry(.4,.2),.8,1,@"thin plus positive chin");
|
|
NumericSettingsCase(c,p,extents[2],Geometry(-.6,-.2),1,-1,@"thin plus negative chin");
|
|
NumericSettingsCase(c,p,extents[0],Geometry(0,0),0,.5,@"half positive chin");
|
|
NumericSettingsCase(c,p,extents[0],Geometry(0,0),0,-.5,@"half negative chin");
|
|
CIImage *image=CoordinateImage(extents[1]);NSData *source=Pixels(c,image);
|
|
PGBeautySettings s=PGBeautyDefaults();s.faceThin=1;s.trackingEnabled=1;NSError *error=nil;
|
|
NSDictionary *g=Geometry(.9,.15);[p setTrackingData:Archive(g) forTarget:0];
|
|
PGFaceFrame *frame=[p processImage:image settings:s time:kCMTimeZero error:&error];
|
|
Check(frame!=nil && [frame.deformationGeometry[0][@"points76"] count]==76,@"offline restoration retains all 76 indexed points");
|
|
PGFaceProcessor *restored=[PGFaceProcessor new];[restored setTrackingData:[p trackingDataForTarget:0] forTarget:0];
|
|
Check([Pixels(c,[restored processImage:image settings:s time:kCMTimeZero error:&error].image) isEqualToData:Pixels(c,frame.image)],@"76-point geometry survives archive round trip with identical pixels");
|
|
NSData *imageWarp=Pixels(c,[p warpImage:frame.colorMask geometry:@[] settings:s extent:image.extent]);
|
|
Check(imageWarp.length==source.length,@"warped support masks retain dimensions");
|
|
// Apply the same coordinate field to an independent scalar mask ramp.
|
|
CIImage *mask=[image imageByApplyingFilter:@"CIColorMatrix" withInputParameters:@{@"inputRVector":[CIVector vectorWithX:1 Y:0 Z:0 W:0],@"inputGVector":[CIVector vectorWithX:1 Y:0 Z:0 W:0],@"inputBVector":[CIVector vectorWithX:1 Y:0 Z:0 W:0]}];
|
|
NSData *maskData=Pixels(c,[p warpImage:mask geometry:@[g] settings:s extent:image.extent]),*frameData=Pixels(c,frame.image);
|
|
const float *m=maskData.bytes,*v=frameData.bytes;
|
|
BOOL masksAgree=YES;for(NSUInteger i=0;i<source.length/sizeof(float);i+=4)if(fabs(m[i]-v[i])>1e-6){masksAgree=NO;break;}
|
|
Check(masksAgree,@"scalar support masks and RGB use the same source coordinates");
|
|
for(NSUInteger mode=0;mode<4;mode++) {
|
|
NSMutableDictionary *bad=[g mutableCopy];
|
|
if(mode==0)[bad removeObjectForKey:@"points76"];
|
|
if(mode==1)bad[@"points76"]=[g[@"points76"] subarrayWithRange:NSMakeRange(0,75)];
|
|
if(mode==2){NSMutableArray *a=[g[@"points76"] mutableCopy];a[65]=@[@(NAN),@.5];bad[@"points76"]=a;}
|
|
if(mode==3){NSMutableArray *a=[g[@"points76"] mutableCopy];a[65]=@[@9,@.5];bad[@"points76"]=a;}
|
|
[p setTrackingData:Archive(bad) forTarget:0];frame=[p processImage:image settings:s time:kCMTimeZero error:&error];
|
|
Check(frame!=nil && frame.deformationGeometry.count==1,@"legacy/malformed indexed points retain the valid face record");
|
|
Check([Pixels(c,frame.image) isEqualToData:source],@"legacy/malformed indexed points never trigger a guessed symmetric thin warp");
|
|
PGBeautySettings chinOnly=s;chinOnly.faceThin=0;chinOnly.faceChin=1;
|
|
Check([Pixels(c,[p processImage:image settings:chinOnly time:kCMTimeZero error:&error].image) isEqualToData:source],@"legacy/malformed indexed points never trigger guessed bbox chin deformation");
|
|
PGBeautySettings other=s;other.faceNarrow=.5;
|
|
Check(![Pixels(c,[p processImage:image settings:other time:kCMTimeZero error:&error].image) isEqualToData:source],@"other controls remain available in legacy archives");
|
|
}
|
|
PGRealtimeFaceTracker *tracker=[PGRealtimeFaceTracker new];PGBeautySettings live=PGBeautyDefaults();
|
|
NSDictionary *a=[tracker resultForGeometry:@[Geometry(0,0)] settings:live time:kCMTimeZero extent:image.extent];
|
|
NSDictionary *b=[tracker resultForGeometry:@[Geometry(.2,0)] settings:live time:CMTimeMake(1,25) extent:image.extent];
|
|
Check([a[@"output"][0][@"points76"] count]==76 && [b[@"output"][0][@"points76"] count]==76,@"realtime stabilization retains the complete indexed point layout");
|
|
Check([b isEqual:[tracker resultForGeometry:@[Geometry(.2,0)] settings:live time:CMTimeMake(1,25) extent:image.extent]],@"repeated rendering does not smooth the 76-point layout again");
|
|
NSDictionary *held=[tracker resultForGeometry:@[] settings:live time:CMTimeMake(2,25) extent:image.extent];
|
|
Check([held[@"output"][0][@"points76"] isEqual:b[@"output"][0][@"points76"]],@"missing detection holds the complete point layout");
|
|
NSDictionary *turned=Geometry(1.4,0);
|
|
NSDictionary *back=[tracker resultForGeometry:@[turned] settings:live time:CMTimeMake(3,25) extent:image.extent];
|
|
Check([back[@"output"][0][@"points76"] isEqual:turned[@"points76"]],@"reacquired side-face points do not blend with a held frontal pose");
|
|
for(int index=1;index<argc;index++) {@autoreleasepool {
|
|
CIImage *portrait=[CIImage imageWithContentsOfURL:[NSURL fileURLWithPath:@(argv[index])] options:@{kCIImageColorSpace:NSNull.null}];
|
|
PGFaceProcessor *real=[PGFaceProcessor new];PGBeautySettings rs=PGBeautyDefaults();rs.faceThin=1;
|
|
PGFaceFrame *f=[real processImage:portrait settings:rs time:kCMTimeZero error:&error];
|
|
Check(f!=nil && f.faces.count>0 && [f.deformationGeometry[0][@"points76"] count]==76,@"real Vision detection supplies exactly 76 points to thin deformation");
|
|
Check(![Pixels(c,f.image) isEqualToData:Pixels(c,portrait)],@"maximum thin changes a real detected portrait");
|
|
Check([real analyzeImage:portrait settings:rs time:kCMTimeZero error:&error],@"real pose analysis saves the complete geometry");
|
|
PGFaceProcessor *copy=[PGFaceProcessor new];[copy setTrackingData:[real trackingDataForTarget:0] forTarget:0];rs.trackingEnabled=1;
|
|
Check([Pixels(c,[copy processImage:portrait settings:rs time:kCMTimeZero error:&error].image) isEqualToData:Pixels(c,f.image)],@"real pose replays from persisted 76 points with identical pixels");
|
|
AnatomicalChecks(c,real,f.deformationGeometry.firstObject,portrait.extent);
|
|
}}
|
|
printf("PASS: %lu slim alignment checks; maximum UV oracle error %.9g\n",(unsigned long)checks,maxOracleError);return 0;
|
|
}}
|