add visualization.py for sanity check

src/trackers.py

@@ -1,6 +1,7 @@
 import torch
 from torch.autograd import Variable
 import numpy as np
+import math
 from sklearn.utils.linear_assignment_ import linear_assignment
 from scipy.optimize import linear_sum_assignment
 from collections import deque
@@ -122,8 +123,14 @@ def similarity(self, bbox, feature=None):
         Computes similarity between the RANTrack and the new detection
         """
 
-        # linear combination of instances in the external memory using alpha
-        pass
+        # compute log probability
+        diff2 = torch.pow(self.mu_bbox - bbox, 2)
+        M = (diff2 / self.sigma_bbox).sum()
+        log_scale = self.sigma_bbox.log().sum()
+
+        constant = math.log(2 * math.pi) * len(bbox)
+
+        return -0.5 * (constant + log_scale + M)
 
 
 class RANTracker(object):

src/visualization.py

@@ -0,0 +1,42 @@
+import os
+import numpy as np
+import time
+from skimage import io
+import matplotlib.pyplot as plt
+import matplotlib.patches as patches
+from utils import load_mot16_gt
+
+file_path = '/scratch0/MOT/MOT16/'
+
+gt, train_seq = load_mot16_gt(file_path)
+
+colours = np.random.rand(32, 3)
+plt.ion()
+fig = plt.figure(figsize=(16, 12))
+
+
+for num in range(0, 50):
+    fn = os.path.join(file_path, 'train', 'MOT16-13', 'img1', '{:06d}.jpg'.format(num+1))
+
+    im = io.imread(fn)
+    ax1 = fig.add_subplot(111, aspect='equal')
+    ax1.imshow(im)
+
+    tracks = gt[0][gt[0]['frame_num'] == num]
+    mask = tracks['track_id'] == 8
+    bboxes = tracks[['x', 'y', 'w', 'h', 'track_id', 'class']][mask]
+
+    # select_bboxes = bboxes[bboxes[:, 4] == 7]
+
+
+    for bbox in bboxes:
+        bbox = np.array(bbox.tolist(), dtype=np.int32)
+        bbox[0:2] -= bbox[2:4] // 2
+        ax1.add_patch(patches.Rectangle((bbox[0], bbox[1]), bbox[2], bbox[3], fill=False, lw=1, ec=colours[bbox[4]%32,:]))
+        ax1.set_adjustable('box-forced')
+        print(bbox[5])
+
+
+    fig.canvas.flush_events()
+    plt.draw()
+    ax1.cla()