-
Notifications
You must be signed in to change notification settings - Fork 1
/
mixup.py
1738 lines (1514 loc) · 80 KB
/
mixup.py
1
import cv2import torchimport osimport randomimport numpy as npimport xml.etree.ElementTree as ETimport xml.dom.minidomimport sysimg_path = '/media/cjh/code/data/train/img_aug_pad+yuan_708'save_path = '/media/cjh/code/data/train/mixupv2'xml_path = '/media/cjh/code/data/train/label_aug_pad+yuan_708'save_xml = '/media/cjh/code/data/train/mixupv2'img_names = os.listdir(img_path)img_num = len(img_names)print("img_num:", img_num)def mixup(): a = 0 for imgname in img_names: a = a + 1 imgpath = img_path + imgname img = cv2.imread(imgpath) img_h, img_w = img.shape[0], img.shape[1] i = random.randint(0, img_num - 1) print('i:', i) add_path = img_path + img_names[i] addimg = cv2.imread(add_path) add_h, add_w = addimg.shape[0], addimg.shape[1] if add_h != img_h or add_w != img_w: print("resize!") addimg = cv2.resize(addimg, (img_w, img_h), interpolation=cv2.INTER_LINEAR) scale_h, scale_w = img_h / add_h, img_w / add_w lam = np.random.beta(20, 20) mixed_img = lam * img + (1 - lam) * addimg save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, mixed_img) print("save_img:", save_img) if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' xmlfile2 = xml_path + img_names[i][:-4] + '.xml' print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) # 读取两张图片的xml tree1 = ET.parse(xmlfile1) tree2 = ET.parse(xmlfile2) doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + ('%d' % a))) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) for obj in tree2.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(int(bbox.find("xmin").text) * scale_w), int(int(bbox.find("ymin").text) * scale_h), int(int(bbox.find("xmax").text) * scale_w), int(int(bbox.find("ymax").text) * scale_h)] objects.append(obj_struct) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() else: # 接下来要对标签进行读取与处理 pass ''' xmlfile1 = xml_path + imgname[:-4] + '.xml' print("xml_1:", xmlfile1) # 读取两张图片的xml tree1 = ET.parse(xmlfile1) doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + ('%d' % a))) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() '''def rand_bbox(size, lamb): w = size[0] h = size[1] cut_rat = np.sqrt(1. - lamb) cut_w = np.int(w * cut_rat) cut_h = np.int(h * cut_rat) cx = np.random.randint(w) cy = np.random.randint(h) bbx1 = np.clip(cx - cut_w // 2, 0, w) bby1 = np.clip(cy - cut_h // 2, 0, h) bbx2 = np.clip(cx + cut_w // 2, 0, w) bby2 = np.clip(cy + cut_h // 2, 0, h) return bbx1, bbx2, bby1, bby2def generate_cutmix_image(image_batch, image_batch_labels, beta): lam = np.random.beta(beta, beta) rand_index = np.random.permutation(len(image_batch)) bbx1, bby1, bbx2, bby2 = rand_bbox(image_batch[0].shape, lam) print("bbx1:", bbx1) print("bby1:", bby1) print("bbx2:", bbx2) print("bby2:", bby2) image_batch_updated = image_batch.copy() image_batch_updated = np.array(image_batch_updated) image_batch = np.array(image_batch)#计算两个(xmin, ymin, xmax, ymax)的IOUdef bbox_iou(box1, box2, GIoU=False, DIoU=False, CIoU=False): b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3] b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3] #print("value1:", max(0, (min(b1_x2, b2_x2) - max(b1_x1, b2_x1)))) #print("value2:", max(0, (min(b1_y2, b2_y2) - max(b1_y1, b2_y1)))) inter = max(0, (min(b1_x2, b2_x2) - max(b1_x1, b2_x1))) * \ max(0, (min(b1_y2, b2_y2) - max(b1_y1, b2_y1))) # Union Area w1, h1 = b1_x2 - b1_x1, b1_y2 - b1_y1 w2, h2 = b2_x2 - b2_x1, b2_y2 - b2_y1 union = (w1 * h1 + 1e-16) + w2 * h2 - inter iou = inter / union # iou if GIoU or DIoU or CIoU: cw = torch.max(b1_x2, b2_x2) - torch.min(b1_x1, b2_x1) # convex (smallest enclosing box) width ch = torch.max(b1_y2, b2_y2) - torch.min(b1_y1, b2_y1) # convex height if GIoU: # Generalized IoU https://arxiv.org/pdf/1902.09630.pdf c_area = cw * ch + 1e-16 # convex area return iou - (c_area - union) / c_area # GIoU if DIoU or CIoU: # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1 # convex diagonal squared c2 = cw ** 2 + ch ** 2 + 1e-16 # centerpoint distance squared rho2 = ((b2_x1 + b2_x2) - (b1_x1 + b1_x2)) ** 2 / 4 + ((b2_y1 + b2_y2) - (b1_y1 + b1_y2)) ** 2 / 4 if DIoU: return iou - rho2 / c2 # DIoU elif CIoU: # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47 v = (4 / math.pi ** 2) * torch.pow(torch.atan(w2 / h2) - torch.atan(w1 / h1), 2) with torch.no_grad(): alpha = v / (1 - iou + v) return iou - (rho2 / c2 + v * alpha) # CIoU return iou'''获取同批次另一张图片的bbox目标，同时需要满足那个bbox目标和这张图片的iou不能太大'''def getbox(xmlfile1, xmlfile2): tree1 = ET.parse(xmlfile1) tree2 = ET.parse(xmlfile2) bbox1_return = [] size1_return = [] bbox2_return = [] bbox2_name_return = [] for obj in tree1.findall("object"): bbox1_temp = [] bbox = obj.find("bndbox") bbox1_temp.append(int(bbox.find("xmin").text)) bbox1_temp.append(int(bbox.find("ymin").text)) bbox1_temp.append(int(bbox.find("xmax").text)) bbox1_temp.append(int(bbox.find("ymax").text)) bbox1_return.append(bbox1_temp) for size in tree1.findall("size"): size1_return.append(int(size.find("width").text)) size1_return.append(int(size.find("height").text)) size1_return.append(int(size.find("depth").text)) for obj in tree2.findall("object"): bbox2_temp = [] bbox = obj.find("bndbox") bbox2_temp.append(int(bbox.find("xmin").text)) bbox2_temp.append(int(bbox.find("ymin").text)) bbox2_temp.append(int(bbox.find("xmax").text)) bbox2_temp.append(int(bbox.find("ymax").text)) bbox2_return.append(bbox2_temp) bbox2_name_return.append(obj.find("name").text) #同时也要把2的名字返回 return size1_return, bbox1_return, bbox2_return, bbox2_name_returndef cutmix1(): a = 0 for imgname in img_names: imgpath = img_path + '/' + imgname print("imgpath:", imgpath) img = cv2.imread(imgpath) #print("img:", img) img_h, img_w = img.shape[0], img.shape[1] i = random.randint(0, img_num - 1) #i = 0 add_path = img_path + '/' + img_names[i] addimg = cv2.imread(add_path) add_h, add_w = addimg.shape[0], addimg.shape[1] if add_h != img_h or add_w != img_w: print("resize!") addimg = cv2.resize(addimg, (img_w, img_h), interpolation=cv2.INTER_LINEAR) scale_h, scale_w = img_h / add_h, img_w / add_w print("scale_h:", scale_h) print("scale_w:", scale_w) #lamb = 0.5 #size = img.shape #bbx1, bby1, bbx2, bby2 = rand_bbox(size, lamb)//变成获取addimg的目标位置 xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' xmlfile2 = xml_path + '/' + img_names[i][:-4] + '.xml' #bbx1, bby1, bbx2, bby2 = getbox(xmlfile1, xmlfile2) size1, bbox1_list, bbox2_list, bbox2_name = getbox(xmlfile1, xmlfile2) len_bbox1 = len(bbox1_list) len_bbox2 = len(bbox2_list) if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) # 读取最原始的待修改图片的所有bbox标签变成list print("objects_here:", objects) '''#用来验证 bbox_show = bbox1_list[0] print("bbox_show:", bbox_show) show = img[bbox_show[1]:bbox_show[3], bbox_show[0]:bbox_show[2]] save_img = '/media/cjh/code/data/mixupv2/show.jpg' cv2.imwrite(save_img, show) break ''' # 接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) # 接下来要分别计算IOU temp_img = img.copy() for i_1 in range(len_bbox2): for j_1 in range(len_bbox1):#当前图片 # 必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1]#要复制的 box_iouvalue_test = bbox_iou(bbox1, bbox2) s_yuan = img_h * img_w s_kuang = (bbox2[2] - bbox2[0]) * (bbox2[3] - bbox2[1]) zhan = s_kuang / s_yuan print("zhan:", zhan) t = 1 #并且要满足这个目标框大小占当前图片要小于六分之一 if ((box_iouvalue_test <= 0) and (bbox2_name[i_1] == 'bird_nest')): #if t: temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] # 这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] obj_struct["bbox"] = [int(int(bbox2[0]) * scale_w), int(int(bbox2[1]) * scale_h), int(int(bbox2[2]) * scale_w), int(int(bbox2[3]) * scale_h)] objects.append(obj_struct) # 只有符合IOU要求的坐标才会被写入 # 把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + '/' + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() # 把图片写入保存 save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) a = a + 1 #下面这些是单张测试的代码，不需要删除掉 ''' if a == 1: print('i:', i) print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) print("list_1:", len(bbox1_list)) print("list_2:", len(bbox2_list)) print("bbox_list1:", bbox1_list) print("bbox_list2:", bbox2_list) print("size:", size1) #这次不一样，这次先创建xml文件，然后再往里面添加标签 if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct)#读取最原始的待修改图片的所有bbox标签变成list #接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) #接下来要分别计算IOU temp_img = img.copy() for i_1 in range(len_bbox2): for j_1 in range(len_bbox1): #必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] box_iouvalue_test = bbox_iou(bbox1, bbox2) if(box_iouvalue_test <= 0): temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] #这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] obj_struct["bbox"] = [int(int(bbox2[0])), int(int(bbox2[1])), int(int(bbox2[2])), int(int(bbox2[3]))] objects.append(obj_struct)#只有符合IOU要求的坐标才会被写入 #把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() #把图片写入保存 save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' bbox1 = bbox1_list[0] bbox2 = bbox2_list[0] box_iouvalue_test = bbox_iou(bbox1, bbox2)#计算出两个目标的iou,必须要iou<=0的时候才能进行扩增 print("box_iouvalue_test:", box_iouvalue_test) if box_iouvalue_test <= 0: temp_img = img.copy() temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' if a == 0: if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' xmlfile2 = xml_path + img_names[i][:-4] + '.xml' print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) # 读取两张图片的xml tree1 = ET.parse(xmlfile1) tree2 = ET.parse(xmlfile2) doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + ('%d' % a))) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) #把这一段改成读取通过验证的bbox for obj in tree2.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(int(bbox.find("xmin").text) * scale_w), int(int(bbox.find("ymin").text) * scale_h), int(int(bbox.find("xmax").text) * scale_w), int(int(bbox.find("ymax").text) * scale_h)] objects.append(obj_struct) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() ''' #print("size:", size)(360, 640, 3)def cutmix_ok(): a = 0 for imgname in img_names: imgpath = img_path + '//' + imgname print("img_path:", imgpath) # img = cv2.imread(imgpath) img = cv2.imdecode(np.fromfile(imgpath, dtype=np.uint8), -1) img_h, img_w = img.shape[0], img.shape[1] #i = random.randint(0, img_num - 1) #i = 1 #跳转到这里 add_path = img_path + '/' + img_names[i] print("add_path:", add_path) # addimg = cv2.imread(add_path) addimg = cv2.imdecode(np.fromfile(add_path, dtype=np.uint8), -1) add_h, add_w = addimg.shape[0], addimg.shape[1] if add_h != img_h or add_w != img_w: print("resize!") addimg = cv2.resize(addimg, (img_w, img_h), interpolation=cv2.INTER_LINEAR) scale_h, scale_w = img_h / add_h, img_w / add_w print("scale_h:", scale_h) print("scale_w:", scale_w) # lamb = 0.5 # size = img.shape # bbx1, bby1, bbx2, bby2 = rand_bbox(size, lamb)//变成获取addimg的目标位置 xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' xmlfile2 = xml_path + '/' + img_names[i][:-4] + '.xml' # bbx1, bby1, bbx2, bby2 = getbox(xmlfile1, xmlfile2) size1, bbox1_list, bbox2_list, bbox2_name = getbox(xmlfile1, xmlfile2) len_bbox1 = len(bbox1_list) len_bbox2 = len(bbox2_list) if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) # 读取最原始的待修改图片的所有bbox标签变成list # 接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) # 接下来要分别计算IOU temp_img = img.copy() for i_1 in range(len_bbox2): for j_1 in range(len_bbox1): # 必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] bbox2_temp = bbox2 bbox2_temp[0] = int(int(bbox2[0]) * scale_w) bbox2_temp[1] = int(int(bbox2[1]) * scale_h) bbox2_temp[2] = int(int(bbox2[2]) * scale_h) bbox2_temp[3] = int(int(bbox2[3]) * scale_h) box_iouvalue_test = bbox_iou(bbox1, bbox2_temp) print("iou:", box_iouvalue_test) if (box_iouvalue_test <= 0): print("hereokokok") # temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] temp_img[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] = addimg[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] # temp_img[int(int(bbox2[1]) * scale_h):int(int(bbox2[3]) * scale_h), int(int(bbox2[0]) * scale_w):int(int(bbox2[2]) * scale_w)] = addimg[int(int(bbox2[1]) * scale_h):int(int(bbox2[3]) * scale_h) , int(int(bbox2[0]) * scale_w):int(int(bbox2[2]) * scale_w)] # 这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] obj_struct["bbox"] = [int(int(bbox2[0]) * scale_w), int(int(bbox2[1]) * scale_h), int(int(bbox2[2]) * scale_w), int(int(bbox2[3]) * scale_h)] objects.append(obj_struct) # 只有符合IOU要求的坐标才会被写入 else: i = i + 1 # 把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + '/' + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() # 把图片写入保存 save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) + '.jpg' # save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) cv2.imencode('.jpg', temp_img)[1].tofile(save_img) # cv2.imwrite(save_img, temp_img) print("save_img:", save_img) break a = a + 1 # 下面这些是单张测试的代码，不需要删除掉 ''' if a == 1: print('i:', i) print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) print("list_1:", len(bbox1_list)) print("list_2:", len(bbox2_list)) print("bbox_list1:", bbox1_list) print("bbox_list2:", bbox2_list) print("size:", size1) #这次不一样，这次先创建xml文件，然后再往里面添加标签 if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct)#读取最原始的待修改图片的所有bbox标签变成list #接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) #接下来要分别计算IOU temp_img = img.copy() for i_1 in range(len_bbox2): for j_1 in range(len_bbox1): #必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] box_iouvalue_test = bbox_iou(bbox1, bbox2) if(box_iouvalue_test <= 0): temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] #这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] obj_struct["bbox"] = [int(int(bbox2[0])), int(int(bbox2[1])), int(int(bbox2[2])), int(int(bbox2[3]))] objects.append(obj_struct)#只有符合IOU要求的坐标才会被写入 #把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() #把图片写入保存 save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' bbox1 = bbox1_list[0] bbox2 = bbox2_list[0] box_iouvalue_test = bbox_iou(bbox1, bbox2)#计算出两个目标的iou,必须要iou<=0的时候才能进行扩增 print("box_iouvalue_test:", box_iouvalue_test) if box_iouvalue_test <= 0: temp_img = img.copy() temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' if a == 0: if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' xmlfile2 = xml_path + img_names[i][:-4] + '.xml' print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) # 读取两张图片的xml tree1 = ET.parse(xmlfile1) tree2 = ET.parse(xmlfile2) doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + ('%d' % a))) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) #把这一段改成读取通过验证的bbox for obj in tree2.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(int(bbox.find("xmin").text) * scale_w), int(int(bbox.find("ymin").text) * scale_h), int(int(bbox.find("xmax").text) * scale_w), int(int(bbox.find("ymax").text) * scale_h)] objects.append(obj_struct) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() ''' # print("size:", size)(360, 640, 3)#zuihou yige cuowu jiushi guanyu nage xunhuandef cutmix_error(): a = 0 for imgname in img_names: imgpath = img_path + '//' + imgname print("img_path:", imgpath) img = cv2.imread(imgpath) #img = cv2.imdecode(np.fromfile(imgpath, dtype=np.uint8), -1) img_h, img_w = img.shape[0], img.shape[1] print("img_h_check:", img_h) #i = random.randint(0, img_num - 1) #i = 1 t = 1 count = 0 while(t != 0): i = random.randint(0, img_num - 1) if i == a:#如果当前图片和准备的是同一张，i = 2 i = i + 1 continue else: pass # 跳转到这里 add_path = img_path + '/' + img_names[i] addimg = cv2.imread(add_path) #addimg = cv2.imdecode(np.fromfile(add_path, dtype=np.uint8), -1) add_h, add_w = addimg.shape[0], addimg.shape[1] if add_h != img_h or add_w != img_w: print("resize!") addimg = cv2.resize(addimg, (img_w, img_h), interpolation=cv2.INTER_LINEAR) scale_h, scale_w = img_h / add_h, img_w / add_w print("scale_h:", scale_h) print("scale_w:", scale_w) xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' xmlfile2 = xml_path + '/' + img_names[i][:-4] + '.xml' size1, bbox1_list, bbox2_list, bbox2_name = getbox(xmlfile1, xmlfile2) len_bbox1 = len(bbox1_list) len_bbox2 = len(bbox2_list) if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' #print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] if count == 0: objects.append(obj_struct) # 读取最原始的待修改图片的所有bbox标签变成list # 接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) # 接下来要分别计算IOU temp_img = img.copy() breakflag = 0 for i_1 in range(len_bbox2): check = 0 for j_1 in range(len_bbox1): # 必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] bbox2_temp = bbox2 bbox2_temp[0] = int(int(bbox2[0]) * scale_w) bbox2_temp[1] = int(int(bbox2[1]) * scale_h) bbox2_temp[2] = int(int(bbox2[2]) * scale_w) bbox2_temp[3] = int(int(bbox2[3]) * scale_h) box_iouvalue_test = bbox_iou(bbox1, bbox2_temp) print("iou:", box_iouvalue_test) if (box_iouvalue_test <= 0): print("hereokokok") print("add_img_h_check:", addimg.shape[0]) print("bbox2_temp_check:", bbox2_temp) # temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] temp_img[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] = addimg[bbox2_temp[1]: \ bbox2_temp[3], \ bbox2_temp[0]: \ bbox2_temp[2]] #temp_img[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] = 255 if a == 1: #[672, 514, 512, 890] print("check_bbox2:", bbox2_temp) print("add_path_check_now:", add_path) #temp_show = addimg[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] temp_img[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] = addimg[ bbox2_temp[1]: \ bbox2_temp[3], \ bbox2_temp[0]: \ bbox2_temp[2]] temp = addimg[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] save_img1 = '/media/cjh/code/data/mixupv2/show.jpg' # save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) # cv2.imencode('.jpg', temp_img)[1].tofile(save_img) #cv2.imwrite(save_img1, temp_img) # temp_img[int(int(bbox2[1]) * scale_h):int(int(bbox2[3]) * scale_h), int(int(bbox2[0]) * scale_w):int(int(bbox2[2]) * scale_w)] = addimg[int(int(bbox2[1]) * scale_h):int(int(bbox2[3]) * scale_h) , int(int(bbox2[0]) * scale_w):int(int(bbox2[2]) * scale_w)] # 这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] ''' obj_struct["bbox"] = [int(int(bbox2[0]) * scale_w), int(int(bbox2[1]) * scale_h), int(int(bbox2[2]) * scale_w), int(int(bbox2[3]) * scale_h)] ''' obj_struct["bbox"] = [bbox2_temp[0], bbox2_temp[1], bbox2_temp[2], bbox2_temp[3]] print("obj_struct_check:", obj_struct["bbox"]) objects.append(obj_struct) # 只有符合IOU要求的坐标才会被写入 t = 0 breakflag = 1 else: count = count + 1 t = 1 i = i + 1 if a == 1: print("t_check:", t) #sys.exit() ''' if breakflag == 1: break ''' if count == 200: break if t == 0: # 把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + '/' + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() # 把图片写入保存 save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) + '.jpg' # save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) #cv2.imencode('.jpg', temp_img)[1].tofile(save_img) cv2.imwrite(save_img, temp_img) print("save_img:", save_img) else: pass print("i:", i) #break a = a + 1 # 下面这些是单张测试的代码，不需要删除掉 ''' if a == 1: print('i:', i) print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) print("list_1:", len(bbox1_list)) print("list_2:", len(bbox2_list)) print("bbox_list1:", bbox1_list) print("bbox_list2:", bbox2_list) print("size:", size1) #这次不一样，这次先创建xml文件，然后再往里面添加标签 if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct)#读取最原始的待修改图片的所有bbox标签变成list #接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) #接下来要分别计算IOU temp_img = img.copy() for i_1 in range(len_bbox2): for j_1 in range(len_bbox1): #必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] box_iouvalue_test = bbox_iou(bbox1, bbox2) if(box_iouvalue_test <= 0): temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] #这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] obj_struct["bbox"] = [int(int(bbox2[0])), int(int(bbox2[1])), int(int(bbox2[2])), int(int(bbox2[3]))] objects.append(obj_struct)#只有符合IOU要求的坐标才会被写入 #把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() #把图片写入保存 save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' bbox1 = bbox1_list[0] bbox2 = bbox2_list[0] box_iouvalue_test = bbox_iou(bbox1, bbox2)#计算出两个目标的iou,必须要iou<=0的时候才能进行扩增 print("box_iouvalue_test:", box_iouvalue_test) if box_iouvalue_test <= 0: temp_img = img.copy() temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' if a == 0: if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' xmlfile2 = xml_path + img_names[i][:-4] + '.xml' print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) # 读取两张图片的xml tree1 = ET.parse(xmlfile1) tree2 = ET.parse(xmlfile2) doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + ('%d' % a))) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) #把这一段改成读取通过验证的bbox for obj in tree2.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(int(bbox.find("xmin").text) * scale_w), int(int(bbox.find("ymin").text) * scale_h), int(int(bbox.find("xmax").text) * scale_w), int(int(bbox.find("ymax").text) * scale_h)] objects.append(obj_struct) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() ''' # print("size:", size)(360, 640, 3)def cutmix(): a = 0 for imgname in img_names: imgpath = img_path + '//' + imgname print("img_path:", imgpath) img = cv2.imread(imgpath) #img = cv2.imdecode(np.fromfile(imgpath, dtype=np.uint8), -1) img_h, img_w = img.shape[0], img.shape[1] #print("img_h_check:", img_h) #i = random.randint(0, img_num - 1) #i = 1 t = 1 count = 0 objects = [] #i = random.randint(0, img_num - 1) while(t == 1): i = random.randint(0, img_num - 1) print("i_here:", i) print("a_here:", a) if i == a:#如果当前图片和准备的是同一张，i = 2 i = i + 1 #continue #else: #pass # 跳转到这里 add_path = img_path + '/' + img_names[i] addimg = cv2.imread(add_path) #addimg = cv2.imdecode(np.fromfile(add_path, dtype=np.uint8), -1) add_h, add_w = addimg.shape[0], addimg.shape[1] if add_h != img_h or add_w != img_w: #print("resize!") addimg = cv2.resize(addimg, (img_w, img_h), interpolation=cv2.INTER_LINEAR) scale_h, scale_w = img_h / add_h, img_w / add_w #print("scale_h:", scale_h) #print("scale_w:", scale_w) xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' xmlfile2 = xml_path + '/' + img_names[i][:-4] + '.xml' size1, bbox1_list, bbox2_list, bbox2_name = getbox(xmlfile1, xmlfile2) len_bbox1 = len(bbox1_list) len_bbox2 = len(bbox2_list) #objects = [] if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + '/' + imgname[:-4] + '.xml' #print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) #objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] if count == 0: print("count:", count) print("add_here") objects.append(obj_struct) # 读取最原始的待修改图片的所有bbox标签变成list print("object_here:", objects) # 接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) # 接下来要分别计算IOU temp_img = img.copy() breakflag = 0 cal = 0 for i_1 in range(len_bbox2): check = 0 for j_1 in range(len_bbox1): # 必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] bbox2_temp = bbox2 bbox2_temp[0] = int(int(bbox2[0]) * scale_w)#xmin bbox2_temp[1] = int(int(bbox2[1]) * scale_h)#ymin bbox2_temp[2] = int(int(bbox2[2]) * scale_w)#xmax bbox2_temp[3] = int(int(bbox2[3]) * scale_h)#ymax box_iouvalue_test = bbox_iou(bbox1, bbox2_temp) print("i_1:", i_1) print("iou:", box_iouvalue_test) ok = 0 if (bbox2_temp[0] > 0) and (bbox2_temp[1] > 0) and (bbox2_temp[2] < size1[0]) and (bbox2_temp[3] < size1[1]): ok = 1 if (box_iouvalue_test <= 0) and (ok == 1):#zhiyou suoyou iou manzu caikeyi check = check + 1 if check == len_bbox1: cal = cal + 1 print("hereokokok") #print("add_img_h_check:", addimg.shape[0]) #print("bbox2_temp_check:", bbox2_temp) # temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] temp_img[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] = addimg[bbox2_temp[1]: \ bbox2_temp[3], \ bbox2_temp[0]: \ bbox2_temp[2]] #temp_img[bbox2_temp[1]:bbox2_temp[3], bbox2_temp[0]:bbox2_temp[2]] = 255 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] ''' obj_struct["bbox"] = [int(int(bbox2[0]) * scale_w), int(int(bbox2[1]) * scale_h), int(int(bbox2[2]) * scale_w), int(int(bbox2[3]) * scale_h)] ''' obj_struct["bbox"] = [bbox2_temp[0], bbox2_temp[1], bbox2_temp[2], bbox2_temp[3]] #print("obj_struct_check:", obj_struct["bbox"]) objects.append(obj_struct) # 只有符合IOU要求的坐标才会被写入 t = 0 print("objects_there:", objects) #breakflag = 1 else: count = count + 1 t = 1 i = i + 1 break if a == 1: print("t_check:", t) #sys.exit() if cal > 0:#说明前面已经拷贝一个了，直接break t = 0 break if count == 100: t = 2 break if t == 0: # 把所有的objects写入 print("len_objects:", len(objects)) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + '/' + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() # 把图片写入保存 save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) + '.jpg' # save_img = save_path + '/' + imgname[:-8] + '_' + ('%d' % a) #cv2.imencode('.jpg', temp_img)[1].tofile(save_img) cv2.imwrite(save_img, temp_img) print("save_img:", save_img) else: pass print("i:", i) #break a = a + 1 # 下面这些是单张测试的代码，不需要删除掉 ''' if a == 1: print('i:', i) print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) print("list_1:", len(bbox1_list)) print("list_2:", len(bbox2_list)) print("bbox_list1:", bbox1_list) print("bbox_list2:", bbox2_list) print("size:", size1) #这次不一样，这次先创建xml文件，然后再往里面添加标签 if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' print("xml_1_here:", xmlfile1) tree1 = ET.parse(xmlfile1) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct)#读取最原始的待修改图片的所有bbox标签变成list #接下来开始创造新的xml doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + '_' + ('%d' % a) + '.jpg')) root.appendChild(nodeframe) nodesize = doc.createElement("size") nodewidth = doc.createElement("width") nodewidth.appendChild(doc.createTextNode(str(size1[0]))) nodeheight = doc.createElement("height") nodeheight.appendChild(doc.createTextNode(str(size1[1]))) nodedepth = doc.createElement("depth") nodedepth.appendChild(doc.createTextNode(str(size1[2]))) nodesize.appendChild(nodewidth) nodesize.appendChild(nodeheight) nodesize.appendChild(nodedepth) root.appendChild(nodesize) #接下来要分别计算IOU temp_img = img.copy() for i_1 in range(len_bbox2): for j_1 in range(len_bbox1): #必须满足2的框要和1的所有框的IOU都满足条件才能赋值 bbox1 = bbox1_list[j_1] bbox2 = bbox2_list[i_1] box_iouvalue_test = bbox_iou(bbox1, bbox2) if(box_iouvalue_test <= 0): temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] #这里要把对应的坐标也添加进xml文件里面 obj_struct = {} obj_struct["name"] = bbox2_name[i_1] obj_struct["bbox"] = [int(int(bbox2[0])), int(int(bbox2[1])), int(int(bbox2[2])), int(int(bbox2[3]))] objects.append(obj_struct)#只有符合IOU要求的坐标才会被写入 #把所有的objects写入 for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() #把图片写入保存 save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' bbox1 = bbox1_list[0] bbox2 = bbox2_list[0] box_iouvalue_test = bbox_iou(bbox1, bbox2)#计算出两个目标的iou,必须要iou<=0的时候才能进行扩增 print("box_iouvalue_test:", box_iouvalue_test) if box_iouvalue_test <= 0: temp_img = img.copy() temp_img[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] = addimg[bbox2[1]:bbox2[3], bbox2[0]:bbox2[2]] save_img = save_path + imgname[:-8] + '_' + ('%d' % a) + '.jpg' cv2.imwrite(save_img, temp_img) print("save_img:", save_img) ''' ''' if a == 0: if imgname != img_names[i]: # 待增强的和加上去的那张图片不是同一张图片 # 接下来要对标签进行读取与处理 xmlfile1 = xml_path + imgname[:-4] + '.xml' xmlfile2 = xml_path + img_names[i][:-4] + '.xml' print("xml_1:", xmlfile1) print("xml_2:", xmlfile2) # 读取两张图片的xml tree1 = ET.parse(xmlfile1) tree2 = ET.parse(xmlfile2) doc = xml.dom.minidom.Document() root = doc.createElement("annotation") doc.appendChild(root) nodeframe = doc.createElement("frame") nodeframe.appendChild(doc.createTextNode(imgname[:-8] + ('%d' % a))) objects = [] for obj in tree1.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(bbox.find("xmin").text), int(bbox.find("ymin").text), int(bbox.find("xmax").text), int(bbox.find("ymax").text)] objects.append(obj_struct) #把这一段改成读取通过验证的bbox for obj in tree2.findall("object"): obj_struct = {} obj_struct["name"] = obj.find("name").text bbox = obj.find("bndbox") obj_struct["bbox"] = [int(int(bbox.find("xmin").text) * scale_w), int(int(bbox.find("ymin").text) * scale_h), int(int(bbox.find("xmax").text) * scale_w), int(int(bbox.find("ymax").text) * scale_h)] objects.append(obj_struct) for obj in objects: nodeobject = doc.createElement("object") nodename = doc.createElement("name") nodebndbox = doc.createElement("bndbox") nodexmin = doc.createElement("xmin") nodeymin = doc.createElement("ymin") nodexmax = doc.createElement("xmax") nodeymax = doc.createElement("ymax") nodename.appendChild(doc.createTextNode(obj["name"])) nodexmin.appendChild(doc.createTextNode(str(obj["bbox"][0]))) nodeymin.appendChild(doc.createTextNode(str(obj["bbox"][1]))) nodexmax.appendChild(doc.createTextNode(str(obj["bbox"][2]))) nodeymax.appendChild(doc.createTextNode(str(obj["bbox"][3]))) nodebndbox.appendChild(nodexmin) nodebndbox.appendChild(nodeymin) nodebndbox.appendChild(nodexmax) nodebndbox.appendChild(nodeymax) nodeobject.appendChild(nodename) nodeobject.appendChild(nodebndbox) root.appendChild(nodeobject) fp = open(save_xml + imgname[:-8] + '_' + ('%d' % a) + '.xml', "w") doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8') fp.close() ''' # print("size:", size)(360, 640, 3)def main(): cutmix()if __name__ == "__main__": main()