Update README.md

Author: Niraj Dev Pandey

README.md

@@ -30,7 +30,43 @@ The code structures looks like below:
   - get_data_from_XML.py # parse Annotations of PASCAL VOC, helper of generator
   
   ```
-
+## Walk-through
+
+The multibox loss is consist of `L1 smooth loss` and `softmax` loss. Let's see how they llok like 
+
+`Arguments`
+    y_true: Ground truth bounding boxes,
+	tensor of shape (?, num_boxes, 4).
+    y_pred: Predicted bounding boxes,
+	tensor of shape (?, num_boxes, 4).
+`Returns`
+    l1_loss: L1-smooth loss, tensor of shape (?, num_boxes).
+`References` - https://arxiv.org/abs/1504.08083
+
+```python
+def _l1_smooth_loss(self, y_true, y_pred):
+	abs_loss = tf.abs(y_true - y_pred)
+	sq_loss = 0.5 * (y_true - y_pred)**2
+	l1_loss = tf.where(tf.less(abs_loss, 1.0), sq_loss, abs_loss - 0.5)
+	return tf.reduce_sum(l1_loss, -1)
+```
+Now let's walk through the `softmax` loss
+
+ `Arguments`
+    y_true: Ground truth targets,
+	tensor of shape (?, num_boxes, num_classes).
+    y_pred: Predicted logits,
+	tensor of shape (?, num_boxes, num_classes).
+`Returns`
+    softmax_loss: Softmax loss, tensor of shape (?, num_boxes).
+
+```python
+def _softmax_loss(self, y_true, y_pred):
+	y_pred = tf.maximum(tf.minimum(y_pred, 1 - 1e-15), 1e-15)
+	softmax_loss = -tf.reduce_sum(y_true * tf.log(y_pred),
+				      axis=-1)
+	return softmax_loss
+```
 ## Resources
 
 dataset can be downloaded from [http://host.robots.ox.ac.uk/pascal/VOC/, use The VOC2007 Challenge in this example