Python 2 compability (thanks to Balint Fabry)

README.md

@@ -3,19 +3,19 @@
 ## Introduction
 
 This repository contains code for *Simple Online and Realtime Tracking with a Deep Association Metric* (Deep SORT).
+We extend the original [SORT](https://github.com/abewley/sort) algorithm to
+integrate appearance information based on a deep appearance descriptor.
 
 ## Dependencies
 
-Tracking code:
+The code is compatible with Python 2.7 and 3. The following dependencies are
+need to run the tracker:
 
 * NumPy
-* sklean (for linear assignment)
-* OpenCV (for visualization)
+* sklean
+* OpenCV
 
-Additionally, feature generation requires:
-
-* TensorFlow
-* tfslim
+Additionally, feature generation requires TensorFlow (>= 1.0).
 
 ## Installation
 

application_util/image_viewer.py

@@ -58,7 +58,7 @@ def view_roi(mat, roi):
         return mat[sy:ey, sx:ex, :]
 
 
-class ImageViewer:
+class ImageViewer(object):
     """An image viewer with drawing routines and video capture capabilities.
 
     Key Bindings:

application_util/visualization.py

@@ -53,9 +53,10 @@ def create_unique_color_uchar(tag, hue_step=0.41):
     return int(255*r), int(255*g), int(255*b)
 
 
-class NoVisualization:
+class NoVisualization(object):
     """
-    TODO(nwojke): Add documentation.
+    A dummy visualization object that loops through all frames in a given
+    sequence to update the tracker without performing any visualization.
     """
 
     def __init__(self, seq_info):
@@ -80,9 +81,9 @@ class NoVisualization:
             self.frame_idx += 1
 
 
-class Visualization:
+class Visualization(object):
     """
-    TODO(nwojke): Add documentation.
+    This class shows tracking output in an OpenCV image viewer.
     """
 
     def __init__(self, seq_info, update_ms):

deep_sort_app.py

@@ -31,7 +31,8 @@ def gather_sequence_info(sequence_dir, detection_file):
         A dictionary of the following sequence information:
 
         * sequence_name: Name of the sequence
-        * image_filanemes: A dictionary that maps frame indices to image filenames.
+        * image_filenames: A dictionary that maps frame indices to image
+          filenames.
         * detections: A numpy array of detections in MOTChallenge format.
         * groundtruth: A numpy array of ground truth in MOTChallenge format.
         * image_size: Image size (height, width).
@@ -163,7 +164,7 @@ def run(sequence_dir, detection_file, output_file, min_confidence,
     results = []
 
     def frame_callback(vis, frame_idx):
-        print("Frame idx", frame_idx)
+        print("Processing frame %05d" % frame_idx)
 
         # Load image and generate detections.
         detections = create_detections(

generate_videos.py

@@ -57,8 +57,6 @@ if __name__ == "__main__":
         sys.exit()
     for sequence_txt in os.listdir(args.result_dir):
         sequence = os.path.splitext(sequence_txt)[0]
-        if sequence != "KITTI-13":
-            continue
         sequence_dir = os.path.join(args.mot_dir, sequence)
         if not os.path.exists(sequence_dir):
             continue