Revert to local dependency to debug training when using maps

EXPERIMENTS/config.json

@@ -5,9 +5,9 @@
     "learning_rate": 0.01, 
     "min_learning_rate": 1e-05,
     "learning_decay_rate": 0.9999,
-    "prediction_horizon": 50,
+    "prediction_horizon": 30,
-    "minimum_history_length": 10,
+    "minimum_history_length": 5,
-    "maximum_history_length": 80,
+    "maximum_history_length": 50,
     "map_encoder": {
         "PEDESTRIAN": {
             "heading_state_index": [2, 3],
@@ -74,7 +74,7 @@
     "dynamic": {
         "PEDESTRIAN": {
             "name": "SingleIntegrator",
-            "distribution": false,
+            "distribution": true,
             "limits": {}
         }
     },

poetry.lock

@@ -3692,7 +3692,7 @@ description = "This repository contains the code for Trajectron++: Dynamically-F
 optional = false
 python-versions = "^3.9,<3.12"
 files = []
-develop = false
+develop = true
 
 [package.dependencies]
 dill = "^0.3.7"
@@ -3719,10 +3719,8 @@ torch = [
 tqdm = "^4.65.0"
 
 [package.source]
-type = "git"
+type = "directory"
-url = "https://git.rubenvandeven.com/security_vision/Trajectron-plus-plus/"
+url = "../Trajectron-plus-plus"
-reference = "HEAD"
-resolved_reference = "4e883511d31c96f0a894afc714fed05e322cea5b"
 
 [[package]]
 name = "tsmoothie"
@@ -3929,4 +3927,4 @@ watchdog = ["watchdog (>=2.3)"]
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.10,<3.12,"
-content-hash = "51023fede550e6a8066ed1689cfedfa01346d994e0dea773476609508dc9ee8b"
+content-hash = "716919f29853fc36b09594af35aa8ca09d4d3ceb7ad9cf54a85f4042569ecf1c"

pyproject.toml

@@ -16,8 +16,8 @@ rewrite_tracks = "trap.tools:rewrite_raw_track_files"
 [tool.poetry.dependencies]
 python = "^3.10,<3.12,"
 
-#trajectron-plus-plus = { path = "../Trajectron-plus-plus/", develop = true }
+trajectron-plus-plus = { path = "../Trajectron-plus-plus/", develop = true }
-trajectron-plus-plus = { git = "https://git.rubenvandeven.com/security_vision/Trajectron-plus-plus/" }
+#trajectron-plus-plus = { git = "https://git.rubenvandeven.com/security_vision/Trajectron-plus-plus/" }
 torch = [
         { version="1.12.1" },
 #       { url = "https://download.pytorch.org/whl/cu113/torch-1.12.1%2Bcu113-cp38-cp38-linux_x86_64.whl", markers = "python_version ~= '3.8' and sys_platform == 'linux'" },

trap/process_data.py

@@ -145,7 +145,7 @@ def process_data(src_dir: Path, dst_dir: Path, name: str, smooth_tracks: bool, c
     print(max_frame_nr)
 
     # separate call so cursor is kept during multiple loops
-    seed(123)
+    # seed(123)
     shuffle(tracks)
 
     dt1 = RollingAverage()

trap/tools.py

@@ -200,7 +200,7 @@ def transition_path_points(path: np.array, t: float):
     # distance = cum_lenghts[-1] * t
     # ts = np.concatenate((np.array([0.]), cum_lenghts / cum_lenghts[-1]))
     # print(cum_lenghts[-1])
-    DRAW_SPEED = 30 # fixed speed (independent of lenght) TODO)) make variable
+    DRAW_SPEED = 35 # fixed speed (independent of lenght) TODO)) make variable
     ts = np.concatenate((np.array([0.]), cum_lenghts / DRAW_SPEED))
     new_path = [path[0]]
 
@@ -235,6 +235,7 @@ def draw_track_predictions(img: cv2.Mat, track: Track, color_index: int, camera:
     # if convert_points:
     #     current_point = convert_points([current_point])[0]
     
+    lines = []
     for pred_i, pred in enumerate(track.predictions):
         pred_coords = pred #cv2.perspectiveTransform(np.array([pred]), inv_H)[0].tolist()
         # line_points = pred_coords
@@ -249,23 +250,15 @@ def draw_track_predictions(img: cv2.Mat, track: Track, color_index: int, camera:
         color = bgr_colors[color_index % len(bgr_colors)]
         color = tuple([int(c*opacity) for c in color])
 
-        
+        line_points = line_points.reshape((-1,1,2))
-        for start, end in zip(line_points[:-1], line_points[1:]):
+        lines.append(line_points)
-        # for ci in range(0, len(pred_coords)):
+
-            # if ci == 0:
+    # draw in a single pass
-            #     # TODO)) prev point
+    cv2.polylines(img, lines, False, color, 2, cv2.LINE_AA)
-            #     # continue
+        # for start, end in zip(line_points[:-1], line_points[1:]):
-            #     start = [int(p) for p in current_point]
+        #     cv2.line(img, start, end, color, 2, lineType=cv2.LINE_AA)
-            #     # start = [int(p) for p in coords[-1]]
+        #     pass
-            #     # start = [0,0]?
+        #     # cv2.circle(img, end, 2, color, 1, lineType=cv2.LINE_AA)
-            #     # print(start)
-            # else:
-            #     start = [int(p) for p in pred_coords[ci-1]]
-            # end = [int(p) for p in pred_coords[ci]]
-            # print(np.rint(start),np.rint(end).tolist())
-            cv2.line(img, start, end, color, 2, lineType=cv2.LINE_AA)
-            pass
-            # cv2.circle(img, end, 2, color, 1, lineType=cv2.LINE_AA)
 
 def draw_trackjectron_history(img: cv2.Mat, track: Track, color_index: int, convert_points: Optional[Callable]):
     if not track.predictor_history:
@@ -301,12 +294,15 @@ def draw_track_projected(img: cv2.Mat, track: Track, color_index: int, camera: C
     
     point_color = bgr_colors[color_index % len(bgr_colors)]
     cv2.circle(img, to_point(history[0]), 3, point_color, 2)
+
+    points = np.rint(history.reshape((-1,1,2))).astype(np.int32)
+    cv2.polylines(img, [points], False, point_color, 1)
     
     for j in range(len(history)-1):
-        a = history[j]
+        # a = history[j]
         b = history[j+1]
 
-        cv2.line(img, to_point(a), to_point(b), point_color, 1)
+        # cv2.line(img, to_point(a), to_point(b), point_color, 1)
         cv2.circle(img, to_point(b), 3, point_color, 2)
 
 

trap/tracker.py

@@ -691,7 +691,7 @@ class Smoother:
         else:
             # "Unlike Kalman filtering, which focuses on predicting and updating the current state using historical measurements, Kalman smoothing enhances the accuracy of past state values"
             # see https://medium.com/@shahalkp1/kalman-smoothing-using-tsmoothie-0175260464e5
-            self.smoother = KalmanSmoother(component='level_trend',  component_noise={'level':0.03, 'season': .02, 'trend':0.04},n_seasons = 2, copy=None)
+            self.smoother = KalmanSmoother(component='level_trend',  component_noise={'level':0.02, 'season': .01, 'trend':0.02},n_seasons = 2, copy=None)