# Single Shot MultiBox Detector(SSD)

* **Single Shot**: this means that the tasks of object localization and classification are done in a single forward pass of the network
* **MultiBox**: this is the name of a technique for bounding box regression developed by Szegedy et al. (we will briefly cover it shortly)
* **Detector**: The network is an object detector that also classifies those detected objects

## Architecture

![Architecture](https://637078585-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MYsi-h_n0zY_8MKKgyu%2Fuploads%2Fgit-blob-faa37985e575cea10e43d7dc7856cd5ffeaf7061%2Fssd.png?alt=media)

## Choosing default boundary boxes

SSD defines a scale value for each feature map layer. Starting from the left, Conv4\_3 detects objects at the smallest scale 0.2 (or 0.1 sometimes) and then increases linearly to the rightmost layer at a scale of 0.9.

![](https://637078585-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MYsi-h_n0zY_8MKKgyu%2Fuploads%2Fgit-blob-ff1cf5ec123eebb583cd302819aebd8197638a49%2Fscale_ssd.png?alt=media)

the scale value with the target aspect ratios, we compute the width and the height of the default boxes. For layers making 6 predictions, SSD starts with 5 target aspect ratios: 1, 2, 3, 1/2 and 1/3. Then the width and the height of the default boxes are calculated as:

![wh](https://637078585-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MYsi-h_n0zY_8MKKgyu%2Fuploads%2Fgit-blob-79f8ab17211e6cee6ec4d6b114e12aeb416250df%2Fssd_wh.png?alt=media)

The center of each default box to $$(\frac{i+0.5}{|f\_{k}|}, \frac{j+0.5}{|f\_{k}|})$$, where $$|f\_{k}|$$ is the size of the k-th square feature map, $$i, j \in \[0, |f\_{k}|]$$.

Default boundary boxes are chosen manually. The criterion for matching a prior and a ground-truth box is IoU (Intersection Over Union), which is also called **Jaccard index**. The more overlap, the better match. The process of matching looks like follows:

```python
for every ground-truth box:
    match the ground-truth box with prior having the biggest the IoU
for every prior:
    ious = IoU(prior, ground_truth_boxes)
    max_iou = max(ious)
    if max_iou > threshold:
        i = argmax(ious)
        match the prior with ground_truth_boxes[i]
```

> YOLO uses k-means clustering on the training dataset to determine those default boundary boxes.

## Try to verify the number of default boxes in SSD300 (the one implemented)

For conv4\_3, conv10\_2 and conv11\_2, we only associate 4 default boxes at each feature map location - omitting aspect ratios of $$\frac{1}{3}$$and 3.

1. Conv4\_3: $$38 \cdot 38 \cdot 4 = 5776$$
2. Conv7: $$19 \cdot 19 \cdot 6 = 2166$$
3. Conv8\_2: $$10 \cdot 10 \cdot 6 = 600$$
4. Conv9\_2: $$5 \cdot 5 \cdot 6 = 150$$
5. Conv10\_2: $$3 \cdot 3 \cdot 4 = 36$$
6. Conv11\_2: $$4$$

Total: $$5776+ 2166 + 600 + 150 + 36 + 4 = 8732$$

## Loss Function

MultiBox's loss function also combined two critical components that made their way into SSD:

* **Confidence Loss**: this measures how confident the network is of the objectness of the computed bounding box. Categorical cross-entropy is used to compute this loss.
* **Location Loss**: this measures how far away the network's predicted bounding boxes are from the ground truth ones from the training set. L1-smooth Norm is used here.

![img](https://cdn-images-1.medium.com/max/1600/1*7h3MObIuV1d0mbSzjvvYjA.png)

$$multiboxLoss = confidenceLoss + \alpha \* locationLoss$$

Where the alpha term helps us in balancing the contribution of the location loss.

## Hard Negative Mining

During training, as most of the bounding boxes will have low IoU and therefore be interpreted as negative training examples, we may end up with a disproportionate amount of negative examples in our training set. Therefore, instead of using all negative predictions, it is advised to **keep a ratio of negative to positive examples of around 3:1**. The reason why you need to keep negative samples is because the network also needs to learn and be explicitly told what constitutes an incorrect detection.

![Example of hard negative mining](https://637078585-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MYsi-h_n0zY_8MKKgyu%2Fuploads%2Fgit-blob-5f26b59079b5c984e0833c632bd83af3075f5d35%2Fhard_negative_mining.png?alt=media)

## Data Augmentation

* Generated additional training examples with patches of the original image at different IoU ratios (e.g. 0.1, 0.3, 0.5, etc.) and random patches as well.
* Each image is also randomly horizontally flipped with a probability of 0.5.