1717
1818import * as tf from '@tensorflow/tfjs-core' ;
1919// tslint:disable-next-line: no-imports-from-dist
20- import { ALL_ENVS , BROWSER_ENVS , describeWithFlags } from '@tensorflow/tfjs-core/dist/jasmine_util' ;
20+ import { ALL_ENVS , BROWSER_ENVS , describeWithFlags } from '@tensorflow/tfjs-core/dist/jasmine_util' ;
2121// tslint:disable-next-line: no-imports-from-dist
22- import { expectArraysClose } from '@tensorflow/tfjs-core/dist/test_util' ;
23- import { MEDIAPIPE_CONNECTED_KEYPOINTS_PAIRS } from '../constants' ;
22+ import { expectArraysClose } from '@tensorflow/tfjs-core/dist/test_util' ;
23+ import { MEDIAPIPE_CONNECTED_KEYPOINTS_PAIRS } from '../constants' ;
2424
2525import * as handPoseDetection from '../index' ;
26- import { getXYPerFrame , KARMA_SERVER , loadImage , loadVideo } from '../shared/test_util' ;
26+ import { getXYPerFrame , KARMA_SERVER , loadImage , loadVideo } from '../shared/test_util' ;
2727
2828// Measured in pixels.
2929const EPSILON_IMAGE = 12 ;
3030// Measured in pixels.
31- const EPSILON_VIDEO = 18 ;
31+ const EPSILON_VIDEO = 22 ;
3232// Measured in meters.
33- const EPSILON_VIDEO_WORLD = 0.01 ;
33+ const EPSILON_VIDEO_WORLD = 0.012 ;
3434
3535// ref:
3636// https://github.com/google/mediapipe/blob/master/mediapipe/python/solutions/hands_test.py
3737const EXPECTED_HAND_KEYPOINTS_PREDICTION = [
3838 [
39- [ 580 , 34 ] , [ 504 , 50 ] , [ 459 , 94 ] , [ 429 , 146 ] , [ 397 , 182 ] , [ 507 , 167 ] ,
39+ [ 580 , 34 ] , [ 504 , 50 ] , [ 459 , 94 ] , [ 429 , 146 ] , [ 397 , 182 ] , [ 507 , 167 ] ,
4040 [ 479 , 245 ] , [ 469 , 292 ] , [ 464 , 330 ] , [ 545 , 180 ] , [ 534 , 265 ] , [ 533 , 319 ] ,
4141 [ 536 , 360 ] , [ 581 , 172 ] , [ 587 , 252 ] , [ 593 , 304 ] , [ 599 , 346 ] , [ 615 , 168 ] ,
4242 [ 628 , 223 ] , [ 638 , 258 ] , [ 648 , 288 ]
4343 ] ,
4444 [
4545 [ 138 , 343 ] , [ 211 , 330 ] , [ 257 , 286 ] , [ 289 , 237 ] , [ 322 , 203 ] , [ 219 , 216 ] ,
46- [ 238 , 138 ] , [ 249 , 90 ] , [ 253 , 51 ] , [ 177 , 204 ] , [ 184 , 115 ] , [ 187 , 60 ] ,
47- [ 185 , 19 ] , [ 138 , 208 ] , [ 131 , 127 ] , [ 124 , 77 ] , [ 117 , 36 ] , [ 106 , 222 ] ,
48- [ 92 , 159 ] , [ 79 , 124 ] , [ 68 , 93 ]
46+ [ 238 , 138 ] , [ 249 , 90 ] , [ 253 , 51 ] , [ 177 , 204 ] , [ 184 , 115 ] , [ 187 , 60 ] ,
47+ [ 185 , 19 ] , [ 138 , 208 ] , [ 131 , 127 ] , [ 124 , 77 ] , [ 117 , 36 ] , [ 106 , 222 ] ,
48+ [ 92 , 159 ] , [ 79 , 124 ] , [ 68 , 93 ]
4949 ]
5050] ;
5151
@@ -66,7 +66,7 @@ describeWithFlags('MediaPipeHands', ALL_ENVS, () => {
6666
6767 // Note: this makes a network request for model assets.
6868 const detector = await handPoseDetection . createDetector (
69- handPoseDetection . SupportedModels . MediaPipeHands , { runtime : 'tfjs' } ) ;
69+ handPoseDetection . SupportedModels . MediaPipeHands , { runtime : 'tfjs' } ) ;
7070 const input : tf . Tensor3D = tf . zeros ( [ 128 , 128 , 3 ] ) ;
7171
7272 const beforeTensors = tf . memory ( ) . numTensors ;
@@ -84,12 +84,12 @@ describeWithFlags('MediaPipeHands', ALL_ENVS, () => {
8484 it ( 'throws error when runtime is not set.' , async ( done ) => {
8585 try {
8686 await handPoseDetection . createDetector (
87- handPoseDetection . SupportedModels . MediaPipeHands ) ;
87+ handPoseDetection . SupportedModels . MediaPipeHands ) ;
8888 done . fail ( 'Loading without runtime succeeded unexpectedly.' ) ;
8989 } catch ( e ) {
9090 expect ( e . message ) . toEqual (
91- `Expect modelConfig.runtime to be either ` +
92- `'tfjs' or 'mediapipe', but got undefined` ) ;
91+ `Expect modelConfig.runtime to be either ` +
92+ `'tfjs' or 'mediapipe', but got undefined` ) ;
9393 done ( ) ;
9494 }
9595 } ) ;
@@ -102,7 +102,7 @@ describeWithFlags('MediaPipeHands static image ', BROWSER_ENVS, () => {
102102
103103 beforeAll ( async ( ) => {
104104 timeout = jasmine . DEFAULT_TIMEOUT_INTERVAL ;
105- jasmine . DEFAULT_TIMEOUT_INTERVAL = 120000 ; // 2mins
105+ jasmine . DEFAULT_TIMEOUT_INTERVAL = 240000 ; // 2mins
106106 image = await loadImage ( 'hands.jpg' , 720 , 382 ) ;
107107 } ) ;
108108
@@ -115,19 +115,19 @@ describeWithFlags('MediaPipeHands static image ', BROWSER_ENVS, () => {
115115
116116 // Note: this makes a network request for model assets.
117117 detector = await handPoseDetection . createDetector (
118- handPoseDetection . SupportedModels . MediaPipeHands ,
119- { runtime : 'tfjs' , modelType : 'lite' } ) ;
118+ handPoseDetection . SupportedModels . MediaPipeHands ,
119+ { runtime : 'tfjs' , modelType : 'lite' } ) ;
120120
121121 const beforeTensors = tf . memory ( ) . numTensors ;
122122
123123 const result = await detector . estimateHands ( image , {
124124 staticImageMode : true
125125 } as handPoseDetection . MediaPipeHandsTfjsEstimationConfig ) ;
126126 const keypoints = result . map (
127- hand => hand . keypoints . map ( keypoint => [ keypoint . x , keypoint . y ] ) ) ;
127+ hand => hand . keypoints . map ( keypoint => [ keypoint . x , keypoint . y ] ) ) ;
128128
129129 expectArraysClose (
130- keypoints , EXPECTED_HAND_KEYPOINTS_PREDICTION , EPSILON_IMAGE ) ;
130+ keypoints , EXPECTED_HAND_KEYPOINTS_PREDICTION , EPSILON_IMAGE ) ;
131131
132132 expect ( tf . memory ( ) . numTensors ) . toEqual ( beforeTensors ) ;
133133
@@ -141,19 +141,19 @@ describeWithFlags('MediaPipeHands static image ', BROWSER_ENVS, () => {
141141
142142 // Note: this makes a network request for model assets.
143143 detector = await handPoseDetection . createDetector (
144- handPoseDetection . SupportedModels . MediaPipeHands ,
145- { runtime : 'tfjs' , modelType : 'full' } ) ;
144+ handPoseDetection . SupportedModels . MediaPipeHands ,
145+ { runtime : 'tfjs' , modelType : 'full' } ) ;
146146
147147 const beforeTensors = tf . memory ( ) . numTensors ;
148148
149149 const result = await detector . estimateHands ( image , {
150150 staticImageMode : true
151151 } as handPoseDetection . MediaPipeHandsTfjsEstimationConfig ) ;
152152 const keypoints = result . map (
153- hand => hand . keypoints . map ( keypoint => [ keypoint . x , keypoint . y ] ) ) ;
153+ hand => hand . keypoints . map ( keypoint => [ keypoint . x , keypoint . y ] ) ) ;
154154
155155 expectArraysClose (
156- keypoints , EXPECTED_HAND_KEYPOINTS_PREDICTION , EPSILON_IMAGE ) ;
156+ keypoints , EXPECTED_HAND_KEYPOINTS_PREDICTION , EPSILON_IMAGE ) ;
157157
158158 expect ( tf . memory ( ) . numTensors ) . toEqual ( beforeTensors ) ;
159159
@@ -174,11 +174,11 @@ describeWithFlags('MediaPipe Hands video ', BROWSER_ENVS, () => {
174174 jasmine . DEFAULT_TIMEOUT_INTERVAL = 120000 ; // 2mins
175175
176176 expected = await fetch ( `${ KARMA_SERVER } )
177- . then ( response => response . json ( ) )
178- . then ( result => getXYPerFrame ( result ) ) ;
177+ . then ( response => response . json ( ) )
178+ . then ( result => getXYPerFrame ( result ) ) ;
179179
180180 expected3D = await fetch ( `${ KARMA_SERVER } )
181- . then ( response => response . json ( ) ) ;
181+ . then ( response => response . json ( ) ) ;
182182 } ) ;
183183
184184 afterAll ( ( ) => {
@@ -189,25 +189,25 @@ describeWithFlags('MediaPipe Hands video ', BROWSER_ENVS, () => {
189189 // Note: this makes a network request for model assets.
190190 const model = handPoseDetection . SupportedModels . MediaPipeHands ;
191191 detector = await handPoseDetection . createDetector (
192- model , { runtime : 'tfjs' , maxHands : 1 } ) ;
192+ model , { runtime : 'tfjs' , maxHands : 1 } ) ;
193193
194194 const result : number [ ] [ ] [ ] = [ ] ;
195195 const result3D : number [ ] [ ] [ ] = [ ] ;
196196
197- const callback = async ( video : HTMLVideoElement , timestamp : number ) :
198- Promise < handPoseDetection . Keypoint [ ] > => {
199- const hands = await detector . estimateHands ( video , null /* config */ ) ;
197+ const callback = async ( video : HTMLVideoElement , timestamp : number ) :
198+ Promise < handPoseDetection . Keypoint [ ] > => {
199+ const hands = await detector . estimateHands ( video , null /* config */ ) ;
200200
201- // maxNumHands is set to 1.
202- result . push ( hands [ 0 ] . keypoints . map ( kp => [ kp . x , kp . y ] ) ) ;
203- result3D . push ( hands [ 0 ] . keypoints3D . map ( kp => [ kp . x , kp . y , kp . z ] ) ) ;
201+ // maxNumHands is set to 1.
202+ result . push ( hands [ 0 ] . keypoints . map ( kp => [ kp . x , kp . y ] ) ) ;
203+ result3D . push ( hands [ 0 ] . keypoints3D . map ( kp => [ kp . x , kp . y , kp . z ] ) ) ;
204204
205- return hands [ 0 ] . keypoints ;
206- } ;
205+ return hands [ 0 ] . keypoints ;
206+ } ;
207207
208208 await loadVideo (
209- 'asl_hand.25fps.mp4' , 25 /* fps */ , callback , expected ,
210- MEDIAPIPE_CONNECTED_KEYPOINTS_PAIRS , 0 /* simulatedInterval unused */ ) ;
209+ 'asl_hand.25fps.mp4' , 25 /* fps */ , callback , expected ,
210+ MEDIAPIPE_CONNECTED_KEYPOINTS_PAIRS , 0 /* simulatedInterval unused */ ) ;
211211
212212 expectArraysClose ( result , expected , EPSILON_VIDEO ) ;
213213 expectArraysClose ( result3D , expected3D , EPSILON_VIDEO_WORLD ) ;
0 commit comments