fix face issue:

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/12256380

modelscope/pipelines/cv/arc_face_recognition_pipeline.py

@@ -48,6 +48,10 @@ class ArcFaceRecognitionPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['img'] = None
+            return rtn_dict
         align_img = result['img']
         face_img = align_img[:, :, ::-1]  # to rgb
         face_img = np.transpose(face_img, axes=(2, 0, 1))
@@ -57,6 +61,8 @@ class ArcFaceRecognitionPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['img'] is None:
+            return {OutputKeys.IMG_EMBEDDING: None}
         img = input['img'].unsqueeze(0)
         emb = self.face_model(img).detach().cpu().numpy()
         emb /= np.sqrt(np.sum(emb**2, -1, keepdims=True))  # l2 norm

modelscope/pipelines/cv/face_attribute_recognition_pipeline.py

@@ -54,9 +54,15 @@ class FaceAttributeRecognitionPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['img'] = None
+            return rtn_dict
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['img'] is None:
+            return {OutputKeys.SCORES: None, OutputKeys.LABELS: None}
         scores = self.fairface(input['img'])
         assert scores is not None
         return {OutputKeys.SCORES: scores, OutputKeys.LABELS: self.map_list}

modelscope/pipelines/cv/face_liveness_ir_pipeline.py

@@ -57,6 +57,10 @@ class FaceLivenessIrPipeline(FaceProcessingBasePipeline):
     def preprocess(self, input: Input) -> Dict[str, Any]:
 
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['input_tensor'] = None
+            return rtn_dict
         orig_img = LoadImage.convert_to_ndarray(input)
         orig_img = orig_img[:, :, ::-1]
         img = super(FaceLivenessIrPipeline,
@@ -70,6 +74,8 @@ class FaceLivenessIrPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['input_tensor'] is None:
+            return {OutputKeys.SCORES: None, OutputKeys.BOXES: None}
         input_feed = {}
         input_feed[
             self.input_node_name[0]] = input['input_tensor'].cpu().numpy()

modelscope/pipelines/cv/face_liveness_xc_pipeline.py

@@ -64,6 +64,10 @@ class FaceLivenessXcPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['input_tensor'] = None
+            return rtn_dict
         img = result['img']
         img = (img - 127.5) * 0.0078125
         img = np.expand_dims(img, 0).copy()
@@ -74,6 +78,8 @@ class FaceLivenessXcPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['input_tensor'] is None:
+            return {OutputKeys.SCORES: None, OutputKeys.BOXES: None}
         input_feed = {}
         input_feed[
             self.input_node_name[0]] = input['input_tensor'].cpu().numpy()

modelscope/pipelines/cv/face_processing_base_pipeline.py

@@ -101,12 +101,14 @@ class FaceProcessingBasePipeline(Pipeline):
             face_lmks = face_lmks.reshape(5, 2)
             align_img, _ = align_face(img, (112, 112), face_lmks)
 
-        result = {}
-        result['img'] = np.ascontiguousarray(align_img)
-        result['scores'] = [scores]
-        result['bbox'] = bboxes
-        result['lmks'] = face_lmks
-        return result
+            result = {}
+            result['img'] = np.ascontiguousarray(align_img)
+            result['scores'] = [scores]
+            result['bbox'] = bboxes
+            result['lmks'] = face_lmks
+            return result
+        else:
+            return None
 
     def align_face_padding(self, img, rect, padding_size=16, pad_pixel=127):
         rect = np.reshape(rect, (-1, 4))

modelscope/pipelines/cv/face_quality_assessment_pipeline.py

@@ -73,6 +73,10 @@ class FaceQualityAssessmentPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['input_tensor'] = None
+            return rtn_dict
         align_img = result['img']
         face_img = align_img[:, :, ::-1]  # to rgb
         face_img = (face_img / 255. - 0.5) / 0.5
@@ -83,12 +87,14 @@ class FaceQualityAssessmentPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['input_tensor'] is None:
+            return {OutputKeys.SCORES: None, OutputKeys.BOXES: None}
         input_feed = {}
         input_feed[
             self.input_node_name[0]] = input['input_tensor'].cpu().numpy()
         result = self.sess.run(self.out_node_name, input_feed=input_feed)
         assert result is not None
-        scores = [result[0][0][0]]
+        scores = [np.mean(result[0][0])]
         boxes = input['bbox'].cpu().numpy()[np.newaxis, :].tolist()
         return {OutputKeys.SCORES: scores, OutputKeys.BOXES: boxes}
 

modelscope/pipelines/cv/face_recognition_onnx_fm_pipeline.py

@@ -66,6 +66,10 @@ class FaceRecognitionOnnxFmPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['input_tensor'] = None
+            return rtn_dict
         align_img = result['img']
         face_img = align_img[:, :, ::-1]  # to rgb
         face_img = (face_img / 255. - 0.5) / 0.5
@@ -76,6 +80,8 @@ class FaceRecognitionOnnxFmPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['input_tensor'] is None:
+            return {OutputKeys.IMG_EMBEDDING: None}
         input_feed = {}
         input_feed[
             self.input_node_name[0]] = input['input_tensor'].cpu().numpy()

modelscope/pipelines/cv/face_recognition_onnx_ir_pipeline.py

@@ -63,6 +63,10 @@ class FaceRecognitionOnnxIrPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['input_tensor'] = None
+            return rtn_dict
         align_img = result['img']
         face_img = align_img[:, :, ::-1]  # to rgb
         face_img = (face_img / 255. - 0.5) / 0.5
@@ -73,6 +77,8 @@ class FaceRecognitionOnnxIrPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['input_tensor'] is None:
+            return {OutputKeys.IMG_EMBEDDING: None}
         input_feed = {}
         input_feed[
             self.input_node_name[0]] = input['input_tensor'].cpu().numpy()

modelscope/pipelines/cv/face_recognition_ood_pipeline.py

@@ -51,6 +51,10 @@ class FaceRecognitionOodPipeline(FaceProcessingBasePipeline):
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['img'] = None
+            return rtn_dict
         align_img = result['img']
         face_img = align_img[:, :, ::-1]  # to rgb
         face_img = np.transpose(face_img, axes=(2, 0, 1))
@@ -60,7 +64,8 @@ class FaceRecognitionOodPipeline(FaceProcessingBasePipeline):
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
-        assert input['img'] is not None
+        if input['img'] is None:
+            return {OutputKeys.IMG_EMBEDDING: None, OutputKeys.SCORES: None}
         img = input['img'].unsqueeze(0)
         output = self.face_model(img)
         emb = output[0].detach().cpu().numpy()

modelscope/pipelines/cv/facial_expression_recognition_pipeline.py

@@ -49,11 +49,16 @@ class FacialExpressionRecognitionPipeline(FaceProcessingBasePipeline):
         ]
 
     def preprocess(self, input: Input) -> Dict[str, Any]:
-        result = super(FacialExpressionRecognitionPipeline,
-                       self).preprocess(input)
+        result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['img'] = None
+            return rtn_dict
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['img'] is None:
+            return {OutputKeys.SCORES: None, OutputKeys.LABELS: None}
         result = self.fer(input)
         assert result is not None
         scores = result[0].tolist()

modelscope/pipelines/cv/facial_landmark_confidence_pipeline.py

@@ -44,12 +44,23 @@ class FacialLandmarkConfidencePipeline(FaceProcessingBasePipeline):
     def preprocess(self, input: Input) -> Dict[str, Any]:
 
         result = super().preprocess(input)
+        if result is None:
+            rtn_dict = {}
+            rtn_dict['img'] = None
+            return rtn_dict
         img = LoadImage.convert_to_ndarray(input)
         img = img[:, :, ::-1]
         result['orig_img'] = img.astype(np.float32)
         return result
 
     def forward(self, input: Dict[str, Any]) -> Dict[str, Any]:
+        if input['img'] is None:
+            return {
+                OutputKeys.SCORES: None,
+                OutputKeys.POSES: None,
+                OutputKeys.KEYPOINTS: None,
+                OutputKeys.BOXES: None
+            }
         result = self.flcm(input)
         assert result is not None
         lms = result[0].reshape(-1, 10).tolist()

tests/pipelines/test_arc_face_recognition.py

@@ -25,8 +25,11 @@ class FaceRecognitionTest(unittest.TestCase, DemoCompatibilityCheck):
             Tasks.face_recognition, model=self.model_id)
         emb1 = face_recognition(img1)[OutputKeys.IMG_EMBEDDING]
         emb2 = face_recognition(img2)[OutputKeys.IMG_EMBEDDING]
-        sim = np.dot(emb1[0], emb2[0])
-        print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
+        if emb1 is None or emb2 is None:
+            print('No Detected Face.')
+        else:
+            sim = np.dot(emb1[0], emb2[0])
+            print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_demo_compatibility(self):

tests/pipelines/test_face_attribute_recognition.py

@@ -29,7 +29,10 @@ class FaceAttributeRecognitionTest(unittest.TestCase):
             Tasks.face_attribute_recognition, model=self.model_id)
         img_path = 'data/test/images/face_recognition_1.png'
         result = fair_face(img_path)
-        self.show_result(img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(img_path, result)
 
 
 if __name__ == '__main__':

tests/pipelines/test_face_liveness_ir.py

@@ -4,6 +4,7 @@ import unittest
 
 import cv2
 
+from modelscope.outputs import OutputKeys
 from modelscope.pipelines import pipeline
 from modelscope.utils.constant import Tasks
 from modelscope.utils.cv.image_utils import draw_face_detection_no_lm_result
@@ -25,13 +26,19 @@ class FaceLivenessIrTest(unittest.TestCase):
     def test_run_modelhub(self):
         face_detection = pipeline(Tasks.face_liveness, model=self.model_id)
         result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_run_default_model(self):
         face_detection = pipeline(Tasks.face_liveness)
         result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
 
 if __name__ == '__main__':

tests/pipelines/test_face_liveness_rgb.py

@@ -4,6 +4,7 @@ import unittest
 
 import cv2
 
+from modelscope.outputs import OutputKeys
 from modelscope.pipelines import pipeline
 from modelscope.utils.constant import Tasks
 from modelscope.utils.cv.image_utils import draw_face_detection_no_lm_result
@@ -25,13 +26,19 @@ class FaceLivenessRgbTest(unittest.TestCase):
     def test_run_modelhub(self):
         face_detection = pipeline(Tasks.face_liveness, model=self.model_id)
         result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
     @unittest.skipUnless(test_level() >= 1, 'skip test in current test level')
     def test_run_default_model(self):
         face_detection = pipeline(Tasks.face_liveness)
         result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
 
 if __name__ == '__main__':

tests/pipelines/test_face_liveness_xc.py

@@ -4,6 +4,7 @@ import unittest
 
 import cv2
 
+from modelscope.outputs import OutputKeys
 from modelscope.pipelines import pipeline
 from modelscope.utils.constant import Tasks
 from modelscope.utils.cv.image_utils import draw_face_detection_no_lm_result
@@ -25,13 +26,19 @@ class FaceLivenessXcTest(unittest.TestCase):
     def test_run_modelhub(self):
         face_detection = pipeline(Tasks.face_liveness, model=self.model_id)
         result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_run_default_model(self):
         face_detection = pipeline(Tasks.face_liveness)
         result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
 
 if __name__ == '__main__':

tests/pipelines/test_face_quality_assessment.py

@@ -4,6 +4,7 @@ import unittest
 
 import cv2
 
+from modelscope.outputs import OutputKeys
 from modelscope.pipelines import pipeline
 from modelscope.utils.constant import Tasks
 from modelscope.utils.cv.image_utils import draw_face_detection_no_lm_result
@@ -14,6 +15,7 @@ class FaceQualityAssessmentTest(unittest.TestCase):
 
     def setUp(self) -> None:
         self.model_id = 'damo/cv_manual_face-quality-assessment_fqa'
+        self.img_path = 'data/test/images/vision_efficient_tuning_test_sunflower.jpg'
         self.img_path = 'data/test/images/face_recognition_1.png'
 
     def show_result(self, img_path, detection_result):
@@ -23,16 +25,22 @@ class FaceQualityAssessmentTest(unittest.TestCase):
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_run_modelhub(self):
-        face_detection = pipeline(
+        face_quality_assessment = pipeline(
             Tasks.face_quality_assessment, model=self.model_id)
-        result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        result = face_quality_assessment(self.img_path)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_run_default_model(self):
-        face_detection = pipeline(Tasks.face_quality_assessment)
-        result = face_detection(self.img_path)
-        self.show_result(self.img_path, result)
+        face_quality_assessment = pipeline(Tasks.face_quality_assessment)
+        result = face_quality_assessment(self.img_path)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(self.img_path, result)
 
 
 if __name__ == '__main__':

tests/pipelines/test_face_recognition_onnx_fm.py

@@ -25,8 +25,11 @@ class FmFaceRecognitionTest(unittest.TestCase, DemoCompatibilityCheck):
             Tasks.face_recognition, model=self.model_id)
         emb1 = face_recognition(img1)[OutputKeys.IMG_EMBEDDING]
         emb2 = face_recognition(img2)[OutputKeys.IMG_EMBEDDING]
-        sim = np.dot(emb1[0], emb2[0])
-        print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
+        if emb1 is None or emb2 is None:
+            print('No Detected Face.')
+        else:
+            sim = np.dot(emb1[0], emb2[0])
+            print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_demo_compatibility(self):

tests/pipelines/test_face_recognition_onnx_ir.py

@@ -25,8 +25,11 @@ class IrFaceRecognitionTest(unittest.TestCase, DemoCompatibilityCheck):
             Tasks.face_recognition, model=self.model_id)
         emb1 = face_recognition(img1)[OutputKeys.IMG_EMBEDDING]
         emb2 = face_recognition(img2)[OutputKeys.IMG_EMBEDDING]
-        sim = np.dot(emb1[0], emb2[0])
-        print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
+        if emb1 is None or emb2 is None:
+            print('No Detected Face.')
+        else:
+            sim = np.dot(emb1[0], emb2[0])
+            print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_demo_compatibility(self):

tests/pipelines/test_face_recognition_ood.py

@@ -24,15 +24,20 @@ class FaceRecognitionOodTest(unittest.TestCase, DemoCompatibilityCheck):
         face_recognition = pipeline(self.task, model=self.model_id)
         result1 = face_recognition(img1)
         emb1 = result1[OutputKeys.IMG_EMBEDDING]
-        score1 = result1[OutputKeys.SCORES][0][0]
 
         result2 = face_recognition(img2)
         emb2 = result2[OutputKeys.IMG_EMBEDDING]
-        score2 = result2[OutputKeys.SCORES][0][0]
 
-        sim = np.dot(emb1[0], emb2[0])
-        print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
-        print(f'OOD score: img1:{score1:.3f}  img2:{score2:.3f}')
+        if emb1 is None or emb2 is None:
+            print('No Detected Face.')
+        else:
+            sim = np.dot(emb1[0], emb2[0])
+
+            score1 = result1[OutputKeys.SCORES][0][0]
+            score2 = result2[OutputKeys.SCORES][0][0]
+
+            print(f'Cos similarity={sim:.3f}, img1:{img1}  img2:{img2}')
+            print(f'OOD score: img1:{score1:.3f}  img2:{score2:.3f}')
 
     @unittest.skipUnless(test_level() >= 0, 'skip test in current test level')
     def test_demo_compatibility(self):

tests/pipelines/test_facial_expression_recognition.py

@@ -29,7 +29,10 @@ class FacialExpressionRecognitionTest(unittest.TestCase):
             Tasks.facial_expression_recognition, model=self.model_id)
         img_path = 'data/test/images/facial_expression_recognition.jpg'
         result = fer(img_path)
-        self.show_result(img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(img_path, result)
 
 
 if __name__ == '__main__':

tests/pipelines/test_facial_landmark_confidence.py

@@ -28,7 +28,10 @@ class FacialLandmarkConfidenceTest(unittest.TestCase):
         flcm = pipeline(Tasks.face_2d_keypoints, model=self.model_id)
         img_path = 'data/test/images/face_recognition_1.png'
         result = flcm(img_path)
-        self.show_result(img_path, result)
+        if result[OutputKeys.SCORES] is None:
+            print('No Detected Face.')
+        else:
+            self.show_result(img_path, result)
 
 
 if __name__ == '__main__':