I recently read a new paper (late 2019) about a one-shot object detector called CenterNet. Apart from this, I'm using Yolo (V3) one-shot detector, and what surprised me is the close similarity between Yolo V1 and CenterNet.

First, both frameworks treat object detection as a regression problem, each of them outputs a tensor that can be seen as a grid with cells (below is an example of an output grid).

Yolo V1 Grid cell

Each cell in this grid predicts an object class, a box offset relative to the cell's position and a box size. The only major difference between Yolo V1 and CenterNet is that Yolo also predicts an object confidence score, that is represented in CenterNet by the class score. Yolo also predicts 2 boxes.

In brief, the tensor at one cell position is Class + B x (Conf + Size + Off) for Yolo V1 and Class + Size + Off for CenterNet.

The training strategy is quite similar too. Only the cell containing the center of a ground truth is responsible for that detection and thus affects the loss. Cells near the ground truths center (base on the distance for CenterNet and IoU for Darknet) have a reduced penalty in the loss (Focal Loss for CenterNet vs and tuned hyper parameter for Yolo).

The loss functions have near the same structure (see above) except that L1 is preferred in CenterNet while Yolo uses L2, among other subtleties.

enter image description here

My point is not that Yolo V1 and CenterNet are the same — there are not — but they are far closer that it appears at first glance.

The problem is that recent papers like CenterNet (CornerNet, ExtremeNet, Triplet CenterNet, MatrixNet) all claim to be "Keypoint-based detector" while they are not so much different than regular "anchor-based" detectors (that are preconditioned regressors in fact).

Instead I think that the biggest difference between Yolo and CenterNet is the backbone that has a bigger resolution for CenterNet (64x64) while Darknet has 7 or 8 only.

My Question is: do you see a major difference between the two concepts that I may have missed and that could explain the performance gap? I understand that new backbones, new loss functions and better resolutions can improve the accuracy but is there a structural difference between the two approaches?

  • 2
    $\begingroup$ First and foremost CenterNet eliminate the need for NMS after detection, whereas the anchor based object detection need NMS after detection $\endgroup$
    – kkr4k
    Mar 10, 2020 at 7:44
  • 1
    $\begingroup$ @kkr4k maybe this is the only important difference? by identifying local maxima in the heatmap, nearby bboxes are de-duplicated automatically. It's in a sense dual to the IOU de-dedupping scheme. $\endgroup$
    – John Jiang
    Oct 17, 2020 at 18:30

1 Answer 1


You've already mentioned some technical differences between the two architectures. I think a key difference is in their "philosophy" and what they are trying to achieve.

CenterNet's abstract points out that they believe bounding-box-based approaches are inefficient as they generate many candidates for a bounding box that needs an NMS phase. Their approach claims to be more efficient and requires 1 step instead of 2.

Another core benefit of keypoint estimation architectures is that they are more flexible than bounding-box-based ones for scenarios where you need something other than a box. For instance, human pose estimation, which they describe in Chapter 4.2. So, I would mark that as a considerable difference.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .