一、NMS是什么？

NMS（non maximum suppression）即非極大值抑制，廣泛應(yīng)用于傳統(tǒng)的特征提取和深度學(xué)習(xí)的目標(biāo)檢測算法中。
NMS原理是通過篩選出局部極大值得到最優(yōu)解。
在2維邊緣提取中體現(xiàn)在提取邊緣輪廓后將一些梯度方向變化率較小的點(diǎn)篩選掉，避免造成干擾。
在三維關(guān)鍵點(diǎn)檢測中也起到重要作用，篩選掉特征中非局部極值。
在目標(biāo)檢測方面如Yolo和RCNN等模型中均有使用，可以將較小分?jǐn)?shù)的輸出框過濾掉,同樣，在三維基于點(diǎn)云的目標(biāo)檢測模型中亦有使用。

---

二、示例

1.opencv示例

查看opencv源碼，可以知道canny算子中使用了nms，即對sobel等梯度計算方法生成的梯度矩陣中的點(diǎn)求取局部極大值。
其計算方法是比較中心點(diǎn)與其鄰域的梯度值，如果為最大值，則保留，不是的話為0。
源碼可見：
Canny算法解析，opencv源碼實現(xiàn)及實例

//讀取圖片Mat img = imread("true.jpg");Mat Grayimg;resize(img, img, Size(400, 600), 0, 0, INTER_LINEAR);cvtColor(img, Grayimg, COLOR_RGB2GRAY); //轉(zhuǎn)為灰度圖Canny(Grayimg, Grayimg, 100, 300, 3);imshow("picture0", img);imshow("picture", Grayimg);waitKey(0);return 0;

2.PCL示例

點(diǎn)云關(guān)鍵點(diǎn)特征提取算法經(jīng)常會使用nms提取極大值點(diǎn)。
如3D SIFT關(guān)鍵點(diǎn)檢測中需要計算尺度空間中像素點(diǎn)的26鄰域的極值點(diǎn)。
算法原理參考：
PCL 3D-SIFT關(guān)鍵點(diǎn)檢測(Z方向梯度約束)

pcl::SIFTKeypoint<pcl::PointXYZ, pcl::PointWithScale> sift; pcl::PointCloud<pcl::PointWithScale> result; sift.setInputCloud(cloud_xyz); pcl::search::KdTree<pcl::PointXYZ>::Ptr tree(new pcl::search::KdTree<pcl::PointXYZ>()); sift.setSearchMethod(tree); sift.setScales(0.01f, 7, 20); sift.setMinimumContrast(0.001f); sift.compute(result);

3.目標(biāo)檢測中nms示例

nms在深度學(xué)習(xí)領(lǐng)域常用于對box的得分進(jìn)行極大值篩選，在rcnn，yolo, pointnet等模型中廣泛使用。
其算法流程大致為：
1：計算所有box的得分。
2：排序，依次與得分高的box的IOU進(jìn)行對比，如果大于設(shè)定的閾值，就刪除該框。
在yolo源代碼detect.py可見：

pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det) conf_thres:置信度即得分score的閾值，yolo為0.25。 iou_thres：重疊度閾值，為0.45 classes：類別數(shù)，可以設(shè)置保留哪一類的box agnostic_nms:是否去除不同類別之間的框,默認(rèn)false max_det:一張圖片中最大識別種類的個數(shù)，默認(rèn)300 def non_max_suppression(prediction, conf_thres=0.25, iou_thres=0.45, classes=None, agnostic=False, multi_label=False,labels=(), max_det=300):"""Runs Non-Maximum Suppression (NMS) on inference resultsReturns:list of detections, on (n,6) tensor per image [xyxy, conf, cls]"""nc = prediction.shape[2] - 5 # number of classesxc = prediction[..., 4] > conf_thres # candidates# Checksassert 0 <= conf_thres <= 1, f'Invalid Confidence threshold {conf_thres}, valid values are between 0.0 and 1.0'assert 0 <= iou_thres <= 1, f'Invalid IoU {iou_thres}, valid values are between 0.0 and 1.0'# Settingsmin_wh, max_wh = 2, 4096 # (pixels) minimum and maximum box width and heightmax_nms = 30000 # maximum number of boxes into torchvision.ops.nms()time_limit = 10.0 # seconds to quit afterredundant = True # require redundant detectionsmulti_label &= nc > 1 # multiple labels per box (adds 0.5ms/img)merge = False # use merge-NMSt = time.time()output = [torch.zeros((0, 6), device=prediction.device)] * prediction.shape[0]for xi, x in enumerate(prediction): # image index, image inference# Apply constraints# x[((x[..., 2:4] < min_wh) | (x[..., 2:4] > max_wh)).any(1), 4] = 0 # width-heightx = x[xc[xi]] # confidence# Cat apriori labels if autolabellingif labels and len(labels[xi]):l = labels[xi]v = torch.zeros((len(l), nc + 5), device=x.device)v[:, :4] = l[:, 1:5] # boxv[:, 4] = 1.0 # confv[range(len(l)), l[:, 0].long() + 5] = 1.0 # clsx = torch.cat((x, v), 0)# If none remain process next imageif not x.shape[0]:continue# Compute confx[:, 5:] *= x[:, 4:5] # conf = obj_conf * cls_conf# Box (center x, center y, width, height) to (x1, y1, x2, y2)box = xywh2xyxy(x[:, :4])# Detections matrix nx6 (xyxy, conf, cls)if multi_label:i, j = (x[:, 5:] > conf_thres).nonzero(as_tuple=False).Tx = torch.cat((box[i], x[i, j + 5, None], j[:, None].float()), 1)else: # best class onlyconf, j = x[:, 5:].max(1, keepdim=True)x = torch.cat((box, conf, j.float()), 1)[conf.view(-1) > conf_thres]# Filter by classif classes is not None:x = x[(x[:, 5:6] == torch.tensor(classes, device=x.device)).any(1)]# Apply finite constraint# if not torch.isfinite(x).all():# x = x[torch.isfinite(x).all(1)]# Check shapen = x.shape[0] # number of boxesif not n: # no boxescontinueelif n > max_nms: # excess boxesx = x[x[:, 4].argsort(descending=True)[:max_nms]] # sort by confidence# Batched NMSc = x[:, 5:6] * (0 if agnostic else max_wh) # classesboxes, scores = x[:, :4] + c, x[:, 4] # boxes (offset by class), scoresi = torchvision.ops.nms(boxes, scores, iou_thres) # NMSif i.shape[0] > max_det: # limit detectionsi = i[:max_det]if merge and (1 < n < 3E3): # Merge NMS (boxes merged using weighted mean)# update boxes as boxes(i,4) = weights(i,n) * boxes(n,4)iou = box_iou(boxes[i], boxes) > iou_thres # iou matrixweights = iou * scores[None] # box weightsx[i, :4] = torch.mm(weights, x[:, :4]).float() / weights.sum(1, keepdim=True) # merged boxesif redundant:i = i[iou.sum(1) > 1] # require redundancyoutput[xi] = x[i]if (time.time() - t) > time_limit:print(f'WARNING: NMS time limit {time_limit}s exceeded')break # time limit exceededreturn output

---

總結(jié)

以上是生活随笔為你收集整理的NMS（非极大值抑制）算法详解与示例的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。