Batán Jinlei

Typical Object detection pipeline has one component for generating proposals for classification. Proposals are nothing but the candidate regions for the object of interest. Most of the approaches employ a sliding window over the feature map and assigns foreground/background scores depending on the features computed in that window. The neighbourhood windows have similar scores to some extent and are considered as candidate regions. This leads to hundreds of proposals. As the proposal generation method should have high recall, we keep loose constraints in this stage. However processing these many proposals all through the classification network is cumbersome. This leads to a technique which filters the proposals based on some criteria ( which we will see soon) called Non-maximum Suppression.

NMS:

Input: A list of Proposal boxes B, corresponding confidence scores S and overlap threshold N.

Output: A list of filtered proposals D.

Algorithm:

IOU calculation is actually used to measure the overlap between two proposals.

Below is the pseudo code of NMS. I have added comments to understand it better.

Now if you observe the algorithm above, the whole filtering process depends on single threshold value. So selection of threshold value is key for performance of the model. However setting this threshold is tricky. Let us see this scenario.

Assume that the overlap threshold N is 0.5. If there is a proposal with 0.51 IOU and has good confidence score, the box will be removed even though the confidence is higher than many other boxes with less IOU. Because of this, if there are two objects side by side, one of them would be eliminated. A proposal with 0.49 IOU is still kept even though its confidence is very low. Of course this is the known issue with any threshold based technique. Now how do we deal with this? Below is an example of such case. Only the proposal with 0.9 is kept and others will be removed. This reduces the precision of the model.

The simple yet efficient way to deal with this case is to use Soft-NMS. The idea is very simple — “instead of completely removing the proposals with high IOU and high confidence, reduce the confidences of the proposals proportional to IOU value”. Now let us apply this idea to the above example. instead of completely removing the proposals with 0.8 score, keep the proposals but reduce their score as shown in the figure below.

As I have mentioned earlier, the scores 0.4 of both the proposals are calculated based on the IOU values. The score calculation is as follows

So this is just one line change in the implementation of NMS algorithm and it increases the precision to a good extent. The figure below shows both the algorithms (NMS and Soft-NMS), which i took from Soft-NMS paper.

These techniques works well for filtering predictions of a single model, What if you have predictions from multiple models? Weighted boxes fusion is a novel method for combining predictions of object detection models. Check out my article to know more about that.

I gave the github links for implementation of NMS and Soft-NMS below in the references. That’s all from this post, thank you so much for being with me.

1. In the ACD application, click Support > Software Updates
2. Click your name in the top right corner of the window
3. Read the licensing agreement and click Accept
4. Click the correct softphone application for your needs:
5. If prompted, save or run the file; otherwise, the download runs automatically

Mutual of Omaha Insurance Company was chartered in 1909 and offers a diversified line of individual and group health and accident insurance products nationwide

While Zigs are very much a visual bait, they are also effective when the water is coloured which proves that the fish are using more than just sight to detect

A German assault at Langemarck, north of Ypres , was checked, but a French counterattack on October 24 yielded little result Thus, the first phase of the battle

In this tutorial, we are going to learn about how to solve the sh : react - scripts : command not found error after running the npm start…

Etizolam is used in the treatment of short term anxiety and insomnia. Etizolam is a benzodiazepine. It works by increasing the action of a chemical messenger (GABA) which suppresses the abnormal and excessive activity of the nerve cells in the brain.