add prototype of trackers.py

src/trackers.py

@@ -0,0 +1,34 @@
+import torch
+from torch.autograd import Variable
+
+use_cuda = torch.cuda.is_available()
+
+
+class RANTracker(object):
+    def __init__(self, bbox, feature=None):
+        self.bbox = bbox
+        self.feature = feature
+        self.h_bbox = 0
+        self.h_feature = 0
+
+        # use np array for this
+        self.external_bbox = 0
+        self.externel_feature = 0
+
+    def update(self, bbox, feature=None):
+        self.bbox = bbox
+        self.feature = feature
+
+    def predict(self, model):
+        bbox = Variable(torch.from_numpy(self.bbox)).cuda()
+        h_bbox = Variable(torch.from_numpy(self.h_bbox)).cuda()
+        #update h, alpha, sigma using RAN
+        alpha_bbox, sigma_bbox, self.h_bbox = model(bbox, h_bbox)
+
+        self.alpha_bbox = alpha_bbox.data.cpu().to_numpy()
+        self.sigma_bbox = sigma_bbox.data.cpu().to_numpy()
+
+    def similarity(self, bbox, feature=None):
+        """
+        Computes similarity between the new detection and this target
+        """