Programming Exercise - Anchor Boxes

[Moutasem_Akkad]

Hi,

How does these numbers saved in yolo_anchors.txt map to 5 Anchor boxes?

0.57273, 0.677385, 1.87446, 2.06253, 3.33843, 5.47434, 7.88282, 3.52778, 9.77052, 9.16828

I am assuming each two consecutive number represent a Height and Width of a box?

[Moutasem_Akkad]

Also, one more question. In the programming exercise, they are using the word Tensor. Is that the same as Vector?

[balaji.ambresh]

Please try help(yolo_head) to view documentation of anchors.
See this for tensor documentation.

[ai_curious]

@Moutasem_Akkad

I am assuming each two consecutive number represent a Height and Width

conceptually, the answer is ‘yes’, though I think this implementation of YOLO typically puts width first, then height, everywhere the two are used.

if your second question is referring to places in the code like this

box_wh = box_wh * anchors_tensor / conv_dims

it is named that way to distinguish between where the anchors are read in from the file and used as a Python variable of type List, such as in preprocess_true_boxes(), from where they have been converted to TensorFlow objects, such as in yolo_head(), which happens here

    # Reshape to batch, height, width, num_anchors, box_params.
    anchors_tensor = K.reshape(K.variable(anchors), [1, 1, 1, num_anchors, 2])

NOTE the reason it is converted from simple List of tuples to the anchors_tensor is so that it has the correct shape to be multiplied by the network’s predicted output for width and height, which is (m,S,S,B,2)

For more about where those anchor box dimensions came from, you might find this helpful

Deriving YOLO anchor boxes

Deriving Anchor Boxes for YOLO One area of recurring questions about YOLO is anchor boxes, introduced with the second 2016 paper, YOLO9000 aka YOLO v2, as an optimization to help improve model stability during training. I’ll defer discussion of the role they play for now and just cover how they are created. Once you see what they are, I think it will be easier to understand how they are used. The Dataset I worked with the Berkeley DeepDrive 100K dataset (BDD). To keep the size manageable, I u…

This thread summarizes how the anchor box shape influences the predicted bounding box width and height at runtime.

Applying YOLO anchor boxes

Applying YOLO Anchor Boxes A previous thread provides an introduction to extracting from labelled ground truth data a small set, on the order of 3 <= k <= 10 , of shapes for what is referred to in the YOLO 9000 paper as dimension clusters or priors, but in this class as anchor boxes . NOTE: Just to make things more confusing, the YOLO 9000 paper uses anchor boxes to refer instead to the hand-picked shapes described in the Faster R-CNN paper, not the ones generated by K-Means. The YOLO 9000 pap…

Hope this helps