corner-net inference过程

inference

 

1. sigmoid 出左上和右下角点的heatmap

tl_heat = torch.sigmoid(tl_heat) br_heat = torch.sigmoid(br_heat)

2. 然后进行一个NMS 也就是最大池化

tl_heat = _nms(tl_heat, kernel=kernel) 
br_heat = _nms(br_heat, kernel=kernel) 

def _nms(heat, kernel=1): 
    pad = (kernel - 1) // 2 
    hmax = nn.functional.max_pool2d(heat, (kernel, kernel), stride=1, padding=pad) 
    keep = (hmax == heat).float() return heat * keep

3. 选择出左上和右下角点的topK

tl_scores, tl_inds, tl_clses, tl_ys, tl_xs = _topk(tl_heat, K=100)
br_scores, br_inds, br_clses, br_ys, br_xs = _topk(br_heat, K=100)

def _topk(scores, K=20):
    batch, cat, height, width = scores.size()

    topk_scores, topk_inds = torch.topk(scores.view(batch, -1), K)

    topk_clses = (topk_inds / (height * width)).int()

    topk_inds = topk_inds % (height * width)
    topk_ys   = (topk_inds / width).int().float()
    topk_xs   = (topk_inds % width).int().float()
    return topk_scores, topk_inds, topk_clses, topk_ys, topk_xs

4. 加上回归后的偏移量

if tl_regr is not None and br_regr is not None:
        tl_regr = _tranpose_and_gather_feat(tl_regr, tl_inds)
        tl_regr = tl_regr.view(batch, K, 1, 2)
        br_regr = _tranpose_and_gather_feat(br_regr, br_inds)
        br_regr = br_regr.view(batch, 1, K, 2)

        tl_xs = tl_xs + tl_regr[..., 0]
        tl_ys = tl_ys + tl_regr[..., 1]
        br_xs = br_xs + br_regr[..., 0]
        br_ys = br_ys + br_regr[..., 1]

 

5. 对左上和右下的角点进行匹配，100*100=10000个proposal，计算embeeding后特征的距离

bboxes = torch.stack((tl_xs, tl_ys, br_xs, br_ys), dim=3)

tl_tag = _tranpose_and_gather_feat(tl_tag, tl_inds)
tl_tag = tl_tag.view(batch, K, 1)
br_tag = _tranpose_and_gather_feat(br_tag, br_inds)
br_tag = br_tag.view(batch, 1, K)
dists  = torch.abs(tl_tag - br_tag)

6. 用类别过滤掉一些错误的proposal

tl_clses = tl_clses.view(batch, K, 1).expand(batch, K, K) 
br_clses = br_clses.view(batch, 1, K).expand(batch, K, K) 
cls_inds = (tl_clses != br_clses)

7. 用embeeding的距离过滤掉一些proposal

dist_inds = (dists > ae_threshold)

8. 通过左上角点的坐标小于右下角点的坐标，过滤掉一些proposal

width_inds = (br_xs < tl_xs) height_inds = (br_ys < tl_ys)

最终得到了box