蚌埠网站制作公司费用,网站开发概述,深圳宝安区网站建设公司,公司推广方法有哪些包括BN层、卷积层、池化层、交叉熵、随机梯度下降法、非极大抑制、k均值聚类等秋招常见的代码实现。
1. BN层
import numpy as npdef batch_norm(outputs, gamma, beta, epsilon1e-6, momentum0.9, running_mean0, running_var1)::param outputs: [B, L]:param gamma: mean:p…包括BN层、卷积层、池化层、交叉熵、随机梯度下降法、非极大抑制、k均值聚类等秋招常见的代码实现。
1. BN层
import numpy as npdef batch_norm(outputs, gamma, beta, epsilon1e-6, momentum0.9, running_mean0, running_var1)::param outputs: [B, L]:param gamma: mean:param beta::param epsilon::return:mean np.mean(outputs, axis(0, 2, 3), keepdimsTrue) # 1, C, H, Wvar np.var(outputs, axis(0,2,3), keepdimsTrue) # 1, C, H, W# mean np.mean(outputs, axis0)# var np.var(outputs, axis0)# 滑动平均用于记录mean和var,用于测试running_mean momentum * running_mean (1-momentum) * meanrunning_var momentum * running_var (1-momentum) * varres gamma * ( outputs - mean ) / np.sqrt(var epsilon) betareturn res, running_mean, running_varif __name__ __main__:outputs np.random.random((16, 64, 8, 8))tmp, _, _ batch_norm(outputs, 1, 1, 1e-6)# print(tmp.shape)mean np.mean(tmp[:, 1, :, :])std np.sqrt(np.var(tmp[:, 1, :, :]))print(mean, std)2. 卷积层
import numpy as npdef conv_forward_naive(x, w, b, conv_param)::param x: [N, C_in, H, W]:param w: [C_out, C_in, k1, k2]:param b: [C_out]:param conv_param:- stride:- pad: the number of pixels that will be used to zero-pad the input:return:- out: (N, C_out, H, W)- cache: (x, w, b, conv_param)out NoneN, C_in, H, W x.shapeC_out, _, k1, k2 w.shapestride, padding conv_param[stride], conv_param[pad]H_out (H-k12*padding) // stride 1W_out (W-k22*padding) // stride 1out np.zeros((N, C_out, H_out, W_out))x_pad np.zeros((N, C_in, H2*padding, W2*padding))x_pad[:, :, padding:paddingH, padding:paddingW] xfor i in range(H_out):for j in range(W_out):x_pad_mask x_pad[:, :, i*stride:i*stridek1, j*stride:j*stridek2]for c in range(C_out):out[:, c, i, j] np.sum(x_pad_mask*w[c, :, :, :], axis(1,2,3))out b[None, :, None, None]cache (x, w, b, conv_param)return out, cache3. maxpooling
import numpy as npdef maxpooling_forward(feature, kernel, stride)::param feature: [N, C, H, W]:param kernel: [k1, k2]:param stride: [s1, s2]:return:N, C, H, W feature.shapek1, k2 kernels1, s2 strideH_out (H - k1) // s1 1W_out (W - k2) // s2 1out np.zeros((N, C, H_out, W_out))for i in range(H_out):for j in range(W_out):feature_mask feature[:, :, i*s1:i*s1k1, j*s2:j*s2k2]out[:, :, i, j] np.max(feature_mask, axis(2,3)) # 注意这里的2,3!!!return out4. cross Entropy
import numpy as npdef cross_entropy(label, outputs, reduceTrue)::param label: B x 1:param outputs: B x c:return: lossloss_list []for i in range(len(label)):y label[i]output outputs[i]sum_exp np.sum([np.exp(k) for k in output])prop np.exp(output[y]) / sum_exploss_list.append(-np.log(prop))if reduce:return np.mean(loss_list)else:return np.sum(loss_list)def softmax(t):return np.exp(t) / np.sum(np.exp(t), axis1, keepdimsTrue)def softmax2(t):return np.exp(t) / np.sum(np.exp(t), axis1, keepdimsTrue)def cross_entropy_2(y, y_, onehotTrue, reduceTrue):y softmax(y)if not onehot:cates y.shapey_ np.eye(cates[-1])[y_]if reduce:return np.mean(-np.sum(y_ * np.log(y), axis1))else:return np.sum(-np.sum(y_ * np.log(y), axis1))if __name__ __main__:outputs [[0.5, 0.5], [0, 1], [1, 0]]label [0, 0, 1]print(cross_entropy(label, outputs, True))print(cross_entropy_2(outputs, label, False))5. sgd
import numpy as np
import random
class MYSGD:def __init__(self, training_data, epochs, batch_size, lr, model):self.training_data training_dataself.epochs epochsself.batch_size batch_sizeself.lr lrself.weight [...]self.bias [...]def run(self):n len(self.training_data)for j in range(self.epochs):random.shuffle(self.training_data)mini_batches [self.training_data[k*self.batch_size: (k1)*self.batch_size]for k in range(n//self.batch_size)]for mini_batch in mini_batches:self.updata(mini_batch)def update(self, mini_batch):nabla_b [np.zeros(b.shape) for b in self.bias]nabla_w [np.zeros(w.shape) for w in self.weight]for x, y in mini_batch:delta_nabla_b, delta_nabla_w self.backprop(x, y)nabla_b [nbdnb for nb, dnb in zip(nabla_b, delta_nabla_b)]nabla_w [nwdnw for nw, dnw in zip(nabla_w, delta_nabla_w)]self.weight [w-(self.eta/len(mini_batch))*nw for w, nw in zip(self.weight, nabla_w)]self.bias [b-(self.eta/len(mini_batch))*nb for b, nb in zip(self.bias, nabla_b)]def backprop(self, x, y):6. nms
import numpy as npdef iou_calculate(bbox1, bbox2, modex1y1x2y2):# 我的x11, y11, x12, y12 bbox1x21, y21, x22, y22 bbox2area1 (y12-y111)*(x12-x111)area2 (y22-y211)*(x22-x211)overlap max(min(y12, y22) - max(y11, y21) 1, 0) * max(min(x12, x22) - max(x11, x21) 1, 0)return overlap / (area2 area1 - overlap 1e-6)def bb_intersection_over_union(boxA, boxB):# 别人的boxA [int(x) for x in boxA]boxB [int(x) for x in boxB]xA max(boxA[0], boxB[0])yA max(boxA[1], boxB[1])xB min(boxA[2], boxB[2])yB min(boxA[3], boxB[3])interArea max(0, xB - xA 1) * max(0, yB - yA 1)boxAArea (boxA[2] - boxA[0] 1) * (boxA[3] - boxA[1] 1)boxBArea (boxB[2] - boxB[0] 1) * (boxB[3] - boxB[1] 1)iou interArea / float(boxAArea boxBArea - interArea)return ioudef nms(outputs, scores, T)::param outputs: bboxes, x1y1x2y2:param scores: confidence of each bbox:param T: threshold:return:# 我的outputs np.array(outputs)[np.argsort(-np.array(scores))]saved [True for _ in range(outputs.shape[0])]for i in range(outputs.shape[0]):if saved[i]:for j in range(i1, outputs.shape[0]):if saved[j]:iou iou_calculate(outputs[i], outputs[j])if iou T:saved[j] Falsescores np.sort(-np.array(scores))return outputs[saved], -scores[saved]# 别人的
def nms_others(bboxes, scores, iou_thresh)::param bboxes: 检测框列表:param scores: 置信度列表:param iou_thresh: IOU阈值:return:x1 bboxes[:, 0]y1 bboxes[:, 1]x2 bboxes[:, 2]y2 bboxes[:, 3]areas (y2 - y1) * (x2 - x1)# 结果列表result []index scores.argsort()[::-1] # 对检测框按照置信度进行从高到低的排序并获取索引# 下面的操作为了安全都是对索引处理while index.size 0:# 当检测框不为空一直循环i index[0]result.append(i) # 将置信度最高的加入结果列表# 计算其他边界框与该边界框的IOUx11 np.maximum(x1[i], x1[index[1:]])y11 np.maximum(y1[i], y1[index[1:]])x22 np.minimum(x2[i], x2[index[1:]])y22 np.minimum(y2[i], y2[index[1:]])w np.maximum(0, x22 - x11 1) # 两个边重叠时,也有1列/行像素点是重叠的h np.maximum(0, y22 - y11 1)overlaps w * hious overlaps / (areas[i] areas[index[1:]] - overlaps)# 只保留满足IOU阈值的索引idx np.where(ious iou_thresh)[0]index index[idx 1] # 处理剩余的边框bboxes, scores bboxes[result], scores[result]return bboxes, scoresdef mynms(bboxes, scores, iou_T):x1 bboxes[:, 0]y1 bboxes[:, 1]x2 bboxes[:, 2]y2 bboxes[:, 3]areas (y2-y11) * (x2-x11)ids np.argsort(scores)[::-1]res []while len(ids) 0:i ids[0]res.append(i)x11 np.maximum(x1[i], x1[ids[1:]])x22 np.minimum(x2[i], x2[ids[1:]])y11 np.maximum(y1[i], y1[ids[1:]])y22 np.minimum(y2[i], y1[ids[1:]])# np.maximum(X,Y,None) : X与Y逐位取最大者. 最少两个参数overlap np.maximum(x22-x111, 0) * np.maximum(y22-y111, 0)iou overlap / (areas[i] areas[ids[1:]] - overlap)ids ids[1:][iouT]return bboxes[res], scores[res]if __name__ __main__:outputs [[10, 10, 20, 20], [15, 15, 25, 25], [9, 15, 25, 13]]scores [0.6, 0.8, 0.7]T 0.1print(nms(outputs, scores, T))print(nms_others(np.array(outputs), np.array(scores), T))print(mynms(np.array(outputs), np.array(scores), T))7. k-means
import numpy as np
import copydef check(clusters_last, clusters_center):# clusters_last.sort()# clusters_center.sort()if len(clusters_last) 0:return Falsefor c1, c2 in zip(clusters_last, clusters_center):if np.linalg.norm(c1 - c2) 0:return Falsereturn Truedef kMeans(data, k)::param data: [n, c]:param k: the number of clusters:return:clusters_last []clusters_center [data[i] for i in range(k)] # random choosedwhile not check(clusters_last, clusters_center):clusters_last copy.deepcopy(clusters_center)clusters [[] for _ in range(k)]for i in range(data.shape[0]):min_dis float(inf)for j, center in enumerate(clusters_center):distance np.linalg.norm(center-data[i])if distance min_dis:min_dis distanceidx jclusters[idx].append(data[i])clusters_center []for i in range(k):clusters_center.append(np.mean(clusters[i], axis0))return clusters_centerdef kMeans2(data, k)::param data: [n, c]:param k: the number of clusters:return:clusters_last []clusters_center copy.deepcopy(data[:k]) # random choosedwhile not check(clusters_last, clusters_center):clusters_last copy.deepcopy(clusters_center)clusters [[] for _ in range(k)]for i in range(data.shape[0]):distance np.linalg.norm(clusters_center - data[i], axis1)idx np.argmin(distance)clusters[idx].append(data[i])clusters_center []for i in range(k):clusters_center.append(np.mean(clusters[i], axis0))clusters_center np.array(clusters_center)return clusters_centerif __name__ __main__:data np.random.random((20, 2))print(kMeans(data, 5))print(kMeans2(data, 5))
