enable smoother

trap/config.py

@@ -195,6 +195,10 @@ frame_emitter_parser.add_argument("--video-src",
                     type=Path,
                     nargs='+',
                     default=lambda: list(Path('../DATASETS/VIRAT_subset_0102x/').glob('*.mp4')))
+frame_emitter_parser.add_argument("--video-offset",
+                    help="Start playback from given frame. Note that when src is an array, this applies to all videos individually.",
+                    default=None,
+                    type=int)
 #TODO: camera as source
 
 frame_emitter_parser.add_argument("--video-loop",

trap/frame_emitter.py

@@ -150,6 +150,12 @@ class FrameEmitter:
             target_frame_duration = 1./fps
             logger.info(f"Emit frames at {fps} fps")
 
+            if self.config.video_offset:
+                logger.info(f"Start at frame {self.config.video_offset}")
+                video.set(cv2.CAP_PROP_POS_FRAMES, self.config.video_offset)
+                i = self.config.video_offset
+
+
             if '-' in video_path.stem:
                 path_stem = video_path.stem[:video_path.stem.rfind('-')]
             else:

trap/prediction_server.py

@@ -122,7 +122,7 @@ class PredictionServer:
             logger.warning("Running on CPU. Specifying --eval_device cuda:0 should dramatically speed up prediction")
 
         if self.config.smooth_predictions:
-            self.smoother = Smoother()
+            self.smoother = Smoother(window_len=4)
         
         context = zmq.Context()
         self.trajectory_socket: zmq.Socket = context.socket(zmq.SUB)

trap/renderer.py

@@ -132,9 +132,10 @@ def decorate_frame(frame: Frame, prediction_frame: Frame, first_time) -> np.arra
     #     new_H = S * self.H * np.linalg.inv(S)
     #     warpedFrame = cv2.warpPerspective(img, new_H, (1000,1000))
     #     cv2.imwrite(str(self.config.output_dir / "orig.png"), warpedFrame)
-
+    cv2.rectangle(img, (0,0), (img.shape[1],25), (0,0,0), -1)
+    
     if not prediction_frame:
-        cv2.putText(img, f"Waiting for prediction...", (20,50), cv2.FONT_HERSHEY_PLAIN, 1, (255,255,0), 1)
+        cv2.putText(img, f"Waiting for prediction...", (20,20), cv2.FONT_HERSHEY_PLAIN, 1, (255,255,0), 1)
         # continue
     else:
         inv_H = np.linalg.pinv(prediction_frame.H)
@@ -151,7 +152,8 @@ def decorate_frame(frame: Frame, prediction_frame: Frame, first_time) -> np.arra
             for ci in range(1, len(coords)):
                 start = [int(p) for p in coords[ci-1]]
                 end = [int(p) for p in coords[ci]]
-                color = (255,255,255) if confirmations[ci] else (100,100,100)
+                # color = (255,255,255) if confirmations[ci] else (100,100,100)
+                color = [100+155*ci/len(coords)]*3
                 cv2.line(img, start, end, color, 2, lineType=cv2.LINE_AA)
 
             if not track.predictions or not len(track.predictions):
@@ -180,17 +182,17 @@ def decorate_frame(frame: Frame, prediction_frame: Frame, first_time) -> np.arra
             cv2.putText(img, f"{track_id} ({(track.history[-1].conf or 0):.2f})", (center[0]+8, center[1]), cv2.FONT_HERSHEY_SIMPLEX, fontScale=.7, thickness=2, color=(0,255,0), lineType=cv2.LINE_AA)
 
        
-    cv2.putText(img, f"{frame.index:06d}", (20,50), cv2.FONT_HERSHEY_PLAIN, 1, (255,255,0), 1)
+    cv2.putText(img, f"{frame.index:06d}", (20,20), cv2.FONT_HERSHEY_PLAIN, 1, (255,255,0), 1)
-    cv2.putText(img, f"{frame.time - first_time:.3f}s", (120,50), cv2.FONT_HERSHEY_PLAIN, 1, (255,255,0), 1)
+    cv2.putText(img, f"{frame.time - first_time:.3f}s", (120,20), cv2.FONT_HERSHEY_PLAIN, 1, (255,255,0), 1)
 
     if prediction_frame:
         # render Δt and Δ frames
-        cv2.putText(img, f"{prediction_frame.index - frame.index}", (90,50), cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
+        cv2.putText(img, f"{prediction_frame.index - frame.index}", (90,20), cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
-        cv2.putText(img, f"{prediction_frame.time - time.time():.2f}s", (200,50), cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
+        cv2.putText(img, f"{prediction_frame.time - time.time():.2f}s", (200,20), cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
-        cv2.putText(img, f"{len(prediction_frame.tracks)} tracks", (500,50), cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
+        cv2.putText(img, f"{len(prediction_frame.tracks)} tracks", (500,20), cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
-        cv2.putText(img, f"h: {np.average([len(t.history or []) for t in prediction_frame.tracks.values()])}", (580, 50),  cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
+        cv2.putText(img, f"h: {np.average([len(t.history or []) for t in prediction_frame.tracks.values()])}", (580,20),  cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
-        cv2.putText(img, f"ph: {np.average([len(t.predictor_history or []) for t in prediction_frame.tracks.values()])}", (660, 50),  cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
+        cv2.putText(img, f"ph: {np.average([len(t.predictor_history or []) for t in prediction_frame.tracks.values()])}", (660,20),  cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
-        cv2.putText(img, f"p: {np.average([len(t.predictions or []) for t in prediction_frame.tracks.values()])}", (740, 50),  cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
+        cv2.putText(img, f"p: {np.average([len(t.predictions or []) for t in prediction_frame.tracks.values()])}", (740,20),  cv2.FONT_HERSHEY_PLAIN, 1, (0,0,255), 1)
 
     return img    
 

trap/tracker.py

@@ -26,6 +26,7 @@ from trap.frame_emitter import DetectionState, Frame, Detection, Track
 
 
 from tsmoothie.smoother import KalmanSmoother, ConvolutionSmoother
+import tsmoothie.smoother
 
 # Detection = [int, int, int, int, float, int]
 # Detections = [Detection]
@@ -322,8 +323,9 @@ def run_tracker(config: Namespace, is_running: Event):
 
 class Smoother:
 
-    def __init__(self):
+    def __init__(self, window_len=2):
-        self.smoother = ConvolutionSmoother(window_len=20, window_type='ones', copy=None)
+        self.smoother = ConvolutionSmoother(window_len=window_len, window_type='ones', copy=None)
+        
     
     def smooth_frame_tracks(self, frame: Frame) -> Frame:
         new_tracks = []