Add Timer for Action (#5438)

* 1. add timer for kpt& action 2. trt support for stgcn

* add parameter description

* make warm_frame configurable

deploy/pphuman/config/infer_cfg.yml

@@ -2,6 +2,7 @@ crop_thresh: 0.5
 attr_thresh: 0.5
 kpt_thresh: 0.2
 visual: True
+warmup_frame: 50
 
 DET:
   model_dir: output_inference/mot_ppyolov3/

deploy/pphuman/pipeline.py

@@ -265,7 +265,7 @@ class PipePredictor(object):
         self.cfg = cfg
         self.output_dir = output_dir
 
-        self.warmup_frame = 1
+        self.warmup_frame = self.cfg['warmup_frame']
         self.pipeline_res = Result()
         self.pipe_timer = PipeTimer()
         self.file_name = None
@@ -469,6 +469,8 @@ class PipePredictor(object):
                 self.pipeline_res.update(attr_res, 'attr')
 
             if self.with_action:
+                if frame_id > self.warmup_frame:
+                    self.pipe_timer.module_time['kpt'].start()
                 kpt_pred = self.kpt_predictor.predict_image(
                     crop_input, visual=False)
                 keypoint_vector, score_vector = translate_to_ori_images(
@@ -478,6 +480,9 @@ class PipePredictor(object):
                     keypoint_vector.tolist(), score_vector.tolist()
                 ] if len(keypoint_vector) > 0 else [[], []]
                 kpt_res['bbox'] = ori_bboxes
+                if frame_id > self.warmup_frame:
+                    self.pipe_timer.module_time['kpt'].end()
+
                 self.pipeline_res.update(kpt_res, 'kpt')
 
                 self.kpt_collector.update(kpt_res,
@@ -487,12 +492,16 @@ class PipePredictor(object):
 
                 action_res = {}
                 if state:
+                    if frame_id > self.warmup_frame:
+                        self.pipe_timer.module_time['action'].start()
                     collected_keypoint = self.kpt_collector.get_collected_keypoint(
                     )  # reoragnize kpt output with ID
                     action_input = parse_mot_keypoint(collected_keypoint,
                                                       self.coord_size)
                     action_res = self.action_predictor.predict_skeleton_with_mot(
                         action_input)
+                    if frame_id > self.warmup_frame:
+                        self.pipe_timer.module_time['action'].end()
                     self.pipeline_res.update(action_res, 'action')
 
                 if self.cfg['visual']:

deploy/python/action_infer.py

@@ -80,7 +80,8 @@ class ActionRecognizer(Detector):
             cpu_threads=cpu_threads,
             enable_mkldnn=enable_mkldnn,
             output_dir=output_dir,
-            threshold=threshold)
+            threshold=threshold,
+            delete_shuffle_pass=True)
 
     def predict(self, repeats=1):
         '''

deploy/python/infer.py

@@ -79,10 +79,11 @@ class Detector(object):
         enable_mkldnn_bfloat16 (bool): whether to turn on mkldnn bfloat16
         output_dir (str): The path of output
         threshold (float): The threshold of score for visualization
+        delete_shuffle_pass (bool): whether to remove shuffle_channel_detect_pass in TensorRT. 
+                                    Used by action model.
     """
 
-    def __init__(
-            self,
+    def __init__(self,
                  model_dir,
                  device='CPU',
                  run_mode='paddle',
@@ -95,7 +96,8 @@ class Detector(object):
                  enable_mkldnn=False,
                  enable_mkldnn_bfloat16=False,
                  output_dir='output',
-            threshold=0.5, ):
+                 threshold=0.5,
+                 delete_shuffle_pass=False):
         self.pred_config = self.set_config(model_dir)
         self.predictor, self.config = load_predictor(
             model_dir,
@@ -110,7 +112,8 @@ class Detector(object):
             trt_calib_mode=trt_calib_mode,
             cpu_threads=cpu_threads,
             enable_mkldnn=enable_mkldnn,
-            enable_mkldnn_bfloat16=enable_mkldnn_bfloat16)
+            enable_mkldnn_bfloat16=enable_mkldnn_bfloat16,
+            delete_shuffle_pass=delete_shuffle_pass)
         self.det_times = Timer()
         self.cpu_mem, self.gpu_mem, self.gpu_util = 0, 0, 0
         self.batch_size = batch_size
@@ -591,7 +594,8 @@ def load_predictor(model_dir,
                    trt_calib_mode=False,
                    cpu_threads=1,
                    enable_mkldnn=False,
-                   enable_mkldnn_bfloat16=False):
+                   enable_mkldnn_bfloat16=False,
+                   delete_shuffle_pass=False):
     """set AnalysisConfig, generate AnalysisPredictor
     Args:
         model_dir (str): root path of __model__ and __params__
@@ -603,6 +607,8 @@ def load_predictor(model_dir,
         trt_opt_shape (int): opt shape for dynamic shape in trt
         trt_calib_mode (bool): If the model is produced by TRT offline quantitative
             calibration, trt_calib_mode need to set True
+        delete_shuffle_pass (bool): whether to remove shuffle_channel_detect_pass in TensorRT. 
+                                    Used by action model.
     Returns:
         predictor (PaddlePredictor): AnalysisPredictor
     Raises:
@@ -673,6 +679,8 @@ def load_predictor(model_dir,
     config.enable_memory_optim()
     # disable feed, fetch OP, needed by zero_copy_run
     config.switch_use_feed_fetch_ops(False)
+    if delete_shuffle_pass:
+        config.delete_pass("shuffle_channel_detect_pass")
     predictor = create_predictor(config)
     return predictor, config