gt_argmax_overlaps=overlaps.argmax(axis=0) ValueError: attempt to get argmax of an empty sequence

最新推荐文章于 2024-10-18 11:13:08 发布

原创最新推荐文章于 2024-10-18 11:13:08 发布 · 2.7k 阅读

2 ·

CC 4.0 BY-SA版权

文章标签：

#FasterRCNN

人工智能同时被 3 个专栏收录

113 篇文章

订阅专栏

深度学习

94 篇文章

订阅专栏

常见问题

35 篇文章

订阅专栏

本文详细解析了在使用Faster-RCNN进行目标检测模型训练过程中遇到的ValueError异常，具体表现为尝试获取空序列的最大值。通过检查XML标注文件的完整性，包括标注数量、图像尺寸及边界框坐标，定位并修复数据问题，确保模型训练顺利进行。

问题描述

用 Faster-RCNN 训练检测模型时，训练到中间过程报如下错误：

gt_argmax_overlaps = overlaps.argmax(axis=0)  
ValueError: attempt to get argmax of an empty sequence

解决方法

写一段代码检查数据问题，可能的问题包括：
1、生成了 xml 标注文件，但标注数量为空
2、检测 xml 中的 width、height 是否为 0
3、检测 size 的 xmin、xmax、ymin、ymax 是否在 [0, width]、[0, height] 范围内

输出异常 xml 路径，将其删除或者修正即可再次训练模型了，训练一切正常。

注：我训练模型用的数据是VOC格式

确定要放弃本次机会？

福利倒计时

: :

立减 ¥

普通VIP年卡可用

立即使用

szZack

关注关注

1
点赞
踩
2

收藏

觉得还不错? 一键收藏
0
评论
分享

复制链接

分享到 QQ

分享到新浪微博

扫一扫
打赏
打赏
打赏举报

举报

专栏目录

使用coco数据集，faster rcnn类方法训练出错解决

小火车~

06-20

4226

问题：在caffe框架下，使用coco数据集进行faster rcnn类方法训练，得到如下错误： File "/data/zn/light_head_rcnn/script/py-RFCN-priv/tools/../lib/rpn/anchor_target_layer.py", line 146, in forward argmax_overlaps = overlaps.argma...

linux下安装及配置mysql5.7 (跟着做不会错信我)

fy_1852003327的博客

05-02

5230

linux下安装及配置mysql5.7 【有问题或错误，请私信我将及时改正；借鉴文章标明出处，谢谢】你安装mysql5.7是不是步骤频繁出错？你是不是用mysql官方提供的压缩包，放至Linux下自己解压安装这些操作？那么这篇将解决你的问题。安装道路千千万，思路变通，条条大路通罗马，看一下我这条道路吧！我的真实感受，太烦了，之后找到一个博客，加上自己理解查阅其他一系列资料，找到这种安装my...

参与评论您还未登录，请先登录后发表或查看评论

ValueError: attempt to get argmax of an empty sequence

热门推荐

hzhj的博客

12-12

2万+

本文针对faster rcnn训练中出现上述问题及解决的方法进行简单记录。报错信息 2019-12-12 04:51:57.002253: W tensorflow/core/framework/op_kernel.cc:1610] Invalid argument: ValueError: attempt to get argmax of an empty sequence Traceba...

gt_argmax_overlaps = overlaps.argmax(axis=0) ValueError: attempt to get argmax of an empty sequence错...

weixin_30794851的博客

08-14

4079

在faster rcnn内进行随机裁剪数据增强，训练一段时间后报错： gt_argmax_overlaps = overlaps.argmax(axis=0) ValueError: attempt to get argmax of an empty sequence 参考博客http://blog.csdn.net/jiajunlee/article/details/50470897知原因是...

错误：errors_impl.InvalidArgumentError: ValueError: attempt to get argmax of an empty sequence的解决方案

03-03

3333

有可能在制作数据集的过程中，某些地方导致xml文件里面的坐标丢失，解决办法有两种，一种是删除掉空坐标的xml文件如果这种类型的xml文件数量较少的情况下，第二种就是检查xml文件然后把丢失的坐标点给添加到xml文件中去。这个错误的解决方案也不一定和我一样，这里的提出只是当作一种参考，可能导致错误的原因多种多样，但是如果后面还是出现了这种错误，一定要仔细检查一下数据集。然后在这次训练的过程中，又遇到了这个错误，这次的错误已经没有办法给pass掉了，因为这个错误会直接导致程序运行中断。

Cannot find reference ‘bbox_overlaps‘ in ‘cython_bbox.py‘

crazyman33的专栏

07-16

384

1. 注释掉 from cython_bbox import bbox_overlaps as bbox_ious。

def get_targets_single(self, gt_bboxes_3d, gt_labels_3d, preds_dict, batch_idx): """Generate training targets for a single sample. Args: gt_bboxes_3d (:obj:`LiDARInstance3DBoxes`): Ground truth gt boxes. gt_labels_3d (torch.Tensor): Labels of boxes. preds_dict (dict): dict of prediction result for a single sample Returns: tuple[torch.Tensor]: Tuple of target including \ the following results in order. - torch.Tensor: classification target. [1, num_proposals] - torch.Tensor: classification weights (mask) [1, num_proposals] - torch.Tensor: regression target. [1, num_proposals, 8] - torch.Tensor: regression weights. [1, num_proposals, 8] - torch.Tensor: iou target. [1, num_proposals] - int: number of positive proposals """ num_proposals = preds_dict["center"].shape[-1] # get pred boxes, carefully ! donot change the network outputs score = copy.deepcopy(preds_dict["heatmap"].detach()) center = copy.deepcopy(preds_dict["center"].detach()) height = copy.deepcopy(preds_dict["height"].detach()) dim = copy.deepcopy(preds_dict["dim"].detach()) rot = copy.deepcopy(preds_dict["rot"].detach()) if "vel" in preds_dict.keys(): vel = copy.deepcopy(preds_dict["vel"].detach()) else: vel = None boxes_dict = self.bbox_coder.decode( score, rot, dim, center, height, vel ) # decode the prediction to real world metric bbox bboxes_tensor = boxes_dict[0]["bboxes"] gt_bboxes_tensor = gt_bboxes_3d.tensor.to(score.device) # each layer should do label assign seperately. if self.auxiliary: num_layer = self.num_decoder_layers else: num_layer = 1 assign_result_list = [] for idx_layer in range(num_layer): bboxes_tensor_layer = bboxes_tensor[ self.num_proposals * idx_layer : self.num_proposals * (idx_layer + 1), : ] score_layer = score[ ..., self.num_proposals * idx_layer : self.num_proposals * (idx_layer + 1), ] if self.train_cfg.assigner.type == "HungarianAssigner3D": assign_result = self.bbox_assigner.assign( bboxes_tensor_layer, gt_bboxes_tensor, gt_labels_3d, score_layer, self.train_cfg, ) elif self.train_cfg.assigner.type == "HeuristicAssigner": assign_result = self.bbox_assigner.assign( bboxes_tensor_layer, gt_bboxes_tensor, None, gt_labels_3d, self.query_labels[batch_idx], ) else: raise NotImplementedError assign_result_list.append(assign_result) # combine assign result of each layer assign_result_ensemble = AssignResult( num_gts=sum([res.num_gts for res in assign_result_list]), gt_inds=torch.cat([res.gt_inds for res in assign_result_list]), max_overlaps=torch.cat([res.max_overlaps for res in assign_result_list]), labels=torch.cat([res.labels for res in assign_result_list]), ) sampling_result = self.bbox_sampler.sample( assign_result_ensemble, bboxes_tensor, gt_bboxes_tensor ) pos_inds = sampling_result.pos_inds neg_inds = sampling_result.neg_inds assert len(pos_inds) + len(neg_inds) == num_proposals # create target for loss computation bbox_targets = torch.zeros([num_proposals, self.bbox_coder.code_size]).to( center.device ) bbox_weights = torch.zeros([num_proposals, self.bbox_coder.code_size]).to( center.device ) ious = assign_result_ensemble.max_overlaps ious = torch.clamp(ious, min=0.0, max=1.0) labels = bboxes_tensor.new_zeros(num_proposals, dtype=torch.long) label_weights = bboxes_tensor.new_zeros(num_proposals, dtype=torch.long) if gt_labels_3d is not None: # default label is -1 labels += self.num_classes # both pos and neg have classification loss, only pos has regression and iou loss if len(pos_inds) > 0: pos_bbox_targets = self.bbox_coder.encode(sampling_result.pos_gt_bboxes) bbox_targets[pos_inds, :] = pos_bbox_targets bbox_weights[pos_inds, :] = 1.0 if gt_labels_3d is None: labels[pos_inds] = 1 else: labels[pos_inds] = gt_labels_3d[sampling_result.pos_assigned_gt_inds] if self.train_cfg.pos_weight <= 0: label_weights[pos_inds] = 1.0 else: label_weights[pos_inds] = self.train_cfg.pos_weight if len(neg_inds) > 0: label_weights[neg_inds] = 1.0 # # compute dense heatmap targets device = labels.device gt_bboxes_3d = torch.cat( [gt_bboxes_3d.gravity_center, gt_bboxes_3d.tensor[:, 3:]], dim=1 ).to(device) grid_size = torch.tensor(self.train_cfg["grid_size"]) pc_range = torch.tensor(self.train_cfg["point_cloud_range"]) voxel_size = torch.tensor(self.train_cfg["voxel_size"]) feature_map_size = ( grid_size[:2] // self.train_cfg["out_size_factor"] ) # [x_len, y_len] heatmap = gt_bboxes_3d.new_zeros( self.num_classes, feature_map_size[1], feature_map_size[0] ) for idx in range(len(gt_bboxes_3d)): width = gt_bboxes_3d[idx][3] length = gt_bboxes_3d[idx][4] width = width / voxel_size[0] / self.train_cfg["out_size_factor"] length = length / voxel_size[1] / self.train_cfg["out_size_factor"] if width > 0 and length > 0: radius = gaussian_radius( (length, width), min_overlap=self.train_cfg["gaussian_overlap"] ) radius = max(self.train_cfg["min_radius"], int(radius)) x, y = gt_bboxes_3d[idx][0], gt_bboxes_3d[idx][1] coor_x = ( (x - pc_range[0]) / voxel_size[0] / self.train_cfg["out_size_factor"] ) coor_y = ( (y - pc_range[1]) / voxel_size[1] / self.train_cfg["out_size_factor"] ) center = torch.tensor( [coor_x, coor_y], dtype=torch.float32, device=device ) center_int = center.to(torch.int32) # original # draw_heatmap_gaussian(heatmap[gt_labels_3d[idx]], center_int, radius) # NOTE: fix draw_heatmap_gaussian( heatmap[gt_labels_3d[idx]], center_int[[1, 0]], radius ) mean_iou = ious[pos_inds].sum() / max(len(pos_inds), 1) return ( labels[None], label_weights[None], bbox_targets[None], bbox_weights[None], ious[None], int(pos_inds.shape[0]), float(mean_iou), heatmap[None], )这些在哪里修改呢？

06-13

raise ValueError(f"Bbox values are out of range [{min_value}, {max_value}]") ``` 此外，还需要确保 `bbox` 的形状和维度与预期一致。可以使用以下代码进行验证： ```python def validate_shape(bbox, expected...

#!/bin/env python # -*- coding: utf-8 -*- ################################################# # Author: songwenhua # Function: MI-13自动出具POFV孔口图纸 # Date: 2025-11-18 # v1.00 songwenhua 用户需求号: 2796 任务ID：2110 # LOAD_MODE__ # /home/incam/Desktop/scripts/danban/auto_make_tuzhi/specific_tuzhi/auto_POFV.py import os import math import re import sys import string import faulthandler faulthandler.enable() from PyQt5.QtGui import * from PyQt5.QtCore import * from PyQt5.QtWidgets import * from PyQt5 import QtWidgets from py39COM import Gateway, InCAM from py39Tools import TableWidget from messageBox import messageBox from ICO import ICO from ICNET import ICNET from reportlab.pdfgen.canvas import Canvas as ReportCanvas from reportlab.platypus import PageBreak, FrameBreak, Frame, Table, TableStyle from reportlab.platypus.doctemplate import SimpleDocTemplate, PageTemplate, NextPageTemplate from DrawingCreate import DrawingTemplate from reportlab.lib import colors from Gerber2SVG import Origin, Feature from Gerber2Canvas import Gerber2Canvas, Dimension, Direct from reportlab.lib.pagesizes import A4 from reportlab.lib.units import mm from EqHelper import EqHelper from params_window import Ui_Form class Dr_POFV_Map(QWidget): A5 = (100*mm, 100*mm) pageSize = (A5[0], A5[1]) # 页面大小 # pageSize = (A4[1], A4[0]) # 页面大小 bgTemp: DrawingTemplate # 画PDF的模板类 doc: SimpleDocTemplate drawingVer = '01' # 版本号 drawingParams = dict() # PDF模板参数 canvas: ReportCanvas # 画布 tmpLays = [] # 需要删除的临时层 sigDimension = {} # 图纸编号 MI-13 drawNo = 'MI-13' # str # 铜厚测量图纸 drawName: str = '孔到孔距离标注图纸' workstation = None dirPath = None # 公共盘路径 filePath = None # 文件路径 datLay = None # dat层 datWork = 'dat_copy_work_lay' # dat复制的工作层 coorArr = [] # 坐标序列 def __init__(self): self.JOB = os.environ.get('JOB', None) self.STEP = os.environ.get('STEP', None) INCAM_DEBUG = os.getenv('INCAM_DEBUG', None) if INCAM_DEBUG == 'yes': self.incam = Gateway() self.JOB = self.incam.job_name self.STEP = self.incam.step_name self.pid = self.incam.pid else: self.incam = InCAM() self.pid = os.getpid() self.ico = ICO(incam=self.incam) self.icNet = ICNET(incam=self.incam) self.jobName = self.ico.SimplifyJobName(jobName=self.JOB) self.dbSite = self.ico.GetDBSite(JOB=self.JOB) self.SITE = self.ico.GetSite(JOB=self.JOB) self.layerMatrix = self.ico.GetLayerMatrix() self.step_list = self.ico.GetStepList() self.workStep = self.ico.GetEditList()[0] #得到edit步骤 basePath = self.ico.GetWorkFilePath(mode='withname') self.imgPath = os.path.join(self.ico.GetWorkFilePath(), 'output', 'laser_image', self.jobName) os.makedirs(self.imgPath, exist_ok=True) self.imgPath = os.path.join(self.imgPath, 'pofv_ring.png') # 如果公盘文件夹不存在，则将pdf存放在“/tmp”下 if not os.path.isdir(basePath) or not os.path.exists(basePath): basePath = '/tmp' self.filePath = os.path.join(basePath, f"{self.jobName}-MI-13-POFV.pdf") # # 网络共享的基础路径（父目录） # base_network_path = '//10.10.80.178/workfile/output/pdf' # job_folder = self.jobName # target_path = os.path.join(base_network_path, job_folder) # 完整目标路径 # basePath = '/tmp' # 默认降级路径 # try: # # 检查基础网络路径是否存在（即 pdf/ 目录） # if not os.path.exists(base_network_path): # raise OSError(f"Base network path does not exist: {base_network_path}") # if not os.path.isdir(base_network_path): # raise OSError(f"Base network path is not a directory: {base_network_path}") # # 尝试创建 jobName 子目录 # os.makedirs(target_path, exist_ok=True) # # 再次确认有写权限 # test_file = os.path.join(target_path, '.test_write') # with open(test_file, 'w') as f: # f.write('test') # os.remove(test_file) # # 如果一切正常，使用网络路径 # basePath = target_path # except Exception as e: # print(f"[WARNING] Cannot use network path: {e}") # messageBox.showDialog( # title='提示', # text=f'无法访问网络路径，将保存至临时目录 (/tmp)。\n错误: {str(e)}', # buttons=['OK'], # defaultButton='OK' # ) # basePath = '/tmp' # # === 设置最终路径 === # self.imgPath = os.path.join(self.ico.GetWorkFilePath(), 'output', 'laser_image', self.jobName) # os.makedirs(self.imgPath, exist_ok=True) # self.imgPath = os.path.join(self.imgPath, 'pofv_ring.png') # self.filePath = os.path.join(basePath, f"{self.jobName}-MI-13-POFV.pdf") self.totalPage = 1 # PDF总页数(外层) # 获取中文用户名 user = self.ico.GetUserName() usList = ICNET.GetCTypeUserInfo('user2CN') self.userCN = user if user in usList: self.userCN = usList[user] self.run() def find_keys_by_start_or_end(self, data, target): """ 查找所有 start 或 end 等于 target 的钻孔层 """ result = [] for key, value in data.items(): if isinstance(value, dict): # 确保是字典 start = value.get('start') end = value.get('end') if start == target or end == target: result.append(key) return result def chk_touch(self, dat_layer, intersect_layer): self.ico.ClearLayer() self.ico.DispWork(dat_layer) # 重置并设置基础过滤器 self.incam.COM("reset_filter_criteria,filter_name=,criteria=all") self.incam.COM("set_filter_type,filter_name=,lines=yes,pads=yes,surfaces=yes,arcs=yes,text=yes") self.incam.COM("set_filter_polarity,filter_name=,positive=yes,negative=yes") # 找到dat层中touch相交层的物体 self.incam.COM(f"sel_ref_feat,layers={intersect_layer},use=filter,mode=touch,pads_as=shape,f_types=line;pad;surface;arc;text,polarity=positive;negative,include_syms=,exclude_syms=") self.incam.COM('get_select_count') selected_features = int(self.incam.COMANS) return selected_features def get_coords(self, feature): """ 从任意图元中快速提取一个坐标点 (x, y) 特别处理 surface 的 orig 字段 """ # 优先返回已有的 cx/cy pad # all_cor = [] if 'cx' in feature and 'cy' in feature: return round(feature['cx'], 3), round(feature['cy'], 3) # 对于 line 等有 x0/y0 的类型 if 'x0' in feature and 'y0' in feature: return round(feature['x0'], 3), round(feature['y0'], 3) # 处理 surface 的 orig if feature.get('type') == 'surface' and isinstance(feature.get('orig'), list): pattern = r'#O[BS]\s+([-\d.]+)\s+([-\d.]+)' for line in feature['orig']: match = re.search(pattern, line) if match: x = float(match.group(1)) y = float(match.group(2)) return round(x, 3), round(y, 3) # 返回第一个有效坐标即可 return None def run(self): self.ico.OpenStep(step=self.workStep, job=self.JOB) site = self.ico.GetSite(self.JOB) layerMatrix = self.ico.GetLayerMatrix() sig_out_list = layerMatrix['sigOutLay'] sm_lay_list = layerMatrix['smAllLay'] drill_through = layerMatrix['drlThrough'] helper = EqHelper(self.incam, self.JOB, self.workStep) pofv_flag = helper.getIsPOFV() if not (site == '301' and pofv_flag is True): return 0 # 存储每层处理结果 layer_results = {} tmp_drill_layer = "drill_final" self.ico.CreateOrEmptyLay([tmp_drill_layer]) for i, sig_layer in enumerate(sig_out_list): sm_layer = sm_lay_list[i] matching_layers = self.find_keys_by_start_or_end(drill_through, sig_layer) temp_intersect = f"int_{sig_layer}" # is_front = i == 0 # 假设第一个为正面 if not matching_layers: continue # 清理并创建相交层 if self.ico.IsLayerExist([temp_intersect]): self.ico.DelLayer([temp_intersect]) self.ico.CreateOrEmptyLay([temp_intersect]) self.ico.ClearAll() self.ico.DispWork(layer=sig_layer) self.ico.DispLayer(layer=sm_layer) self.ico.GetLayIntersect(self.workStep, sm_layer, sig_layer, acc=0.01) self.incam.COM(f"matrix_rename_layer,job={self.JOB},matrix=matrix,layer=intersect,new_name={temp_intersect}") self.ico.ClearLayer() self.ico.DispWork(temp_intersect) inter_info = self.ico.GetFeatureFullInfo(self.workStep, layer=temp_intersect) if not inter_info: messageBox.showDialog( title='提示', text=f'{sm_layer} 和 {sig_layer} 没有相交部分', buttons=['OK'], defaultButton='OK' ) self.ico.DelLayer(temp_intersect) continue #开始检测该层是否有有效钻孔匹配 result = { 'has_full_match': False, # 缩小50um后仍匹配 'has_raw_match': False, # 原始匹配 'has_expanded_match': False, # 外扩50um后匹配 'dat_layer': None, 'temp_intersect': temp_intersect } found_in_this_layer = False for dat_layer in matching_layers: # 情况A：检查原始是否 touch selected_features = self.chk_touch(dat_layer, temp_intersect) if selected_features > 0: # 尝试缩小50um shrunk_layer = temp_intersect + '-100' self.ico.ClearLayer() self.ico.DispWork(temp_intersect) self.incam.COM( 'copy_layer, source_job = %s, source_step = %s, source_layer = %s, dest = layer_name, ' 'dest_step =, dest_layer = %s, mode = replace, invert = no, copy_notes = no, ' 'copy_attrs = new_layers_only, copy_sr_feat = no' % ( self.JOB, self.workStep, temp_intersect, shrunk_layer) ) self.ico.ClearLayer() # 缩小50um self.ico.DispWork(shrunk_layer) self.incam.COM("rv_tab_empty,report=resize_rep,is_empty=yes") self.incam.COM("sel_resize,size=-100,corner_ctl=no") self.incam.COM("rv_tab_view_results_enabled,report=resize_rep,is_enabled=no,serial_num=-1,all_count=-1") selected_shrunk = self.chk_touch(dat_layer, shrunk_layer) if selected_shrunk > 0: result['has_full_match'] = True result['dat_layer'] = dat_layer self.ico.DelLayer(shrunk_layer) found_in_this_layer = True break # 成功即退出 dat_layer 循环 else: result['has_raw_match'] = True result['dat_layer'] = dat_layer self.ico.DelLayer(shrunk_layer) #情况B：原始无 touch，尝试外扩+50um else: expanded_layer = temp_intersect + '+100' self.ico.ClearLayer() self.ico.DispWork(temp_intersect) self.incam.COM( 'copy_layer, source_job = %s, source_step = %s, source_layer = %s, dest = layer_name, ' 'dest_step =, dest_layer = %s, mode = replace, invert = no, copy_notes = no, ' 'copy_attrs = new_layers_only, copy_sr_feat = no' % ( self.JOB, self.workStep, temp_intersect, expanded_layer) ) self.ico.ClearLayer() # 外扩50um self.ico.DispWork(expanded_layer) self.incam.COM("rv_tab_empty,report=resize_rep,is_empty=yes") self.incam.COM("sel_resize,size=+100,corner_ctl=no") self.incam.COM("rv_tab_view_results_enabled,report=resize_rep,is_enabled=no,serial_num=-1,all_count=-1") selected_expanded = self.chk_touch(dat_layer, expanded_layer) if selected_expanded > 0: result['has_expanded_match'] = True result['dat_layer'] = dat_layer result['expanded_layer'] = expanded_layer # 保留用于后续复制 found_in_this_layer = True break # 成功即退出 else: self.ico.DelLayer(expanded_layer) # 保存当前层结果 if found_in_this_layer or result['has_raw_match']: layer_results[sig_layer] = result else: self.ico.DelLayer(temp_intersect) # 无任何匹配，清理 # 二、根据收集结果进行最终输出决策 final_copied = False # 1. 优先：正面缩小50um后有匹配 front_sig = sig_out_list[0] if front_sig in layer_results: res = layer_results[front_sig] if res['has_full_match']: temp = res['temp_intersect'] dat = res['dat_layer'] shrunk = temp + '-100' # 重建并使用缩小层 self.ico.ClearLayer() self.ico.DispWork(temp) self.incam.COM( 'copy_layer, source_job = %s, source_step = %s, source_layer = %s, dest = layer_name, ' 'dest_step =, dest_layer = %s, mode = replace, , invert = no, copy_notes = no, ' 'copy_attrs = new_layers_only, copy_sr_feat = no' % (self.JOB, self.workStep, temp, shrunk)) self.incam.COM("sel_resize,size=-100,corner_ctl=no") self.chk_touch(dat, shrunk) self.incam.COM(f"sel_copy_other,dest=layer_name,target_layer={tmp_drill_layer},invert=no,dx=0,dy=0,size=0,x_anchor=-1.36612,y_anchor=-1.03115,subsystem=1-Up-Edit") self.ico.DelLayer(shrunk) final_copied = True # 2. 背面缩小50um后有匹配 if not final_copied: for i, sig_layer in enumerate(sig_out_list): if i == 0: continue # 跳过正面 if sig_layer in layer_results: res = layer_results[sig_layer] if res['has_full_match']: temp = res['temp_intersect'] dat = res['dat_layer'] shrunk = temp + '-100' self.ico.ClearLayer() self.ico.DispWork(temp) self.incam.COM( 'copy_layer, source_job = %s, source_step = %s, source_layer = %s, dest = layer_name, ' 'dest_step =, dest_layer = %s, mode = replace, invert = no, copy_notes = no, ' 'copy_attrs = new_layers_only, copy_sr_feat = no' % ( self.JOB, self.workStep, temp, shrunk_layer)) self.incam.COM("sel_resize,size=-100,corner_ctl=no") self.chk_touch(dat, shrunk) self.incam.COM(f"sel_copy_other,dest=layer_name,target_layer={tmp_drill_layer},invert=no,dx=0,dy=0,size=0,x_anchor=-1.36612,y_anchor=-1.03115,subsystem=1-Up-Edit") self.ico.DelLayer(shrunk) final_copied = True break # 3. 正面有原始或外扩匹配（但没有进入缩小成功分支） if not final_copied and front_sig in layer_results: res = layer_results[front_sig] temp = res['temp_intersect'] dat = res['dat_layer'] self.ico.ClearLayer() self.ico.DispWork(temp) if res['has_raw_match']: self.chk_touch(dat, temp) self.incam.COM(f"sel_copy_other,dest=layer_name,target_layer={tmp_drill_layer},invert=no,dx=0,dy=0,size=0,x_anchor=-1.36612,y_anchor=-1.03115,subsystem=1-Up-Edit") final_copied = True elif res['has_expanded_match']: exp_layer = temp + '+100' # self.incam.COM('copy_layer, ..., dest_layer=%s' % exp_layer) self.incam.COM( 'copy_layer, source_job = %s, source_step = %s, source_layer = %s, dest = layer_name, ' 'dest_step =, dest_layer = %s, mode = replace, invert = no, copy_notes = no, ' 'copy_attrs = new_layers_only, copy_sr_feat = no' % ( self.JOB, self.workStep, temp, exp_layer)) self.incam.COM("sel_resize,size=+100,corner_ctl=no") self.chk_touch(dat, exp_layer) self.incam.COM(f"sel_copy_other,dest=layer_name,target_layer={tmp_drill_layer},invert=no,dx=0,dy=0,size=0,x_anchor=-1.36612,y_anchor=-1.03115,subsystem=1-Up-Edit") self.ico.DelLayer(exp_layer) final_copied = True # 4. 任意背面有原始或外扩匹配 if not final_copied: for i, sig_layer in enumerate(sig_out_list): if i == 0: continue if sig_layer in layer_results: res = layer_results[sig_layer] temp = res['temp_intersect'] dat = res['dat_layer'] self.ico.ClearLayer() self.ico.DispWork(temp) if res['has_raw_match']: self.chk_touch(dat, temp) elif res['has_expanded_match']: exp_layer = temp + '+100' self.incam.COM( 'copy_layer, source_job = %s, source_step = %s, source_layer = %s, dest = layer_name, ' 'dest_step =, dest_layer = %s, mode = replace, invert = no, copy_notes = no, ' 'copy_attrs = new_layers_only, copy_sr_feat = no' % ( self.JOB, self.workStep, temp, exp_layer)) self.incam.COM("sel_resize,size=+100,corner_ctl=no") self.chk_touch(dat, exp_layer) self.ico.DelLayer(exp_layer) self.incam.COM(f"sel_copy_other,dest=layer_name,target_layer={tmp_drill_layer},invert=no,dx=0,dy=0,size=0,x_anchor=-1.36612,y_anchor=-1.03115,subsystem=1-Up-Edit") final_copied = True break # 5. 完全失败 if not final_copied: messageBox.showDialog( title='提示', text='正面和背面均未找到符合条件的钻孔', buttons=['OK'], defaultButton='OK' ) # 清理所有临时层 for sig_layer in sig_out_list: base = f"int_{sig_layer}" self.ico.DelLayer([base, base+'+100']) self.sigDimension[sig_layer] = {} self.sigDimension[sig_layer]['point_x'] = [] self.sigDimension[sig_layer]['point_y'] = [] self.ico.ClearLayer() # 获取要标注的孔坐标（来自 tmp_drill_layer） self.ico.DispWork(tmp_drill_layer) #获取最终孔层的信息，主要是要其中任意一个孔的中心坐标 padList = self.incam.INFO( '-t layer -e %s/%s/%s -m script -d FEATURES -o consider_origin+feat_index+f0' % ( self.JOB, self.workStep, tmp_drill_layer)) for pad in padList: pad.strip() # 3 #P 0.927 1.915 r261 P 1 0 N;.drill=via,.drill_flag=103,.combined_size=0.000000 strList = pad.split() match1 = re.search(r'#(\d+)\s+#P\s+', pad) if match1: midpointX = '%0.3f' % ( float(strList[2])) # 孔盘中点的X坐标 midpointY = '%0.3f' % ( float(strList[3])) # 孔盘中点的Y坐标 self.sigDimension[sig_layer]['point_x'].append(midpointX) self.sigDimension[sig_layer]['point_y'].append(midpointY) self.ico.ClearLayer() # self.ico.DispWork(sig_layer) # self.ico.DispLayer(sm_layer) # self.ico.DispLayer(dat_layer) self.__renderPDF(sig_layer) #在找到的那一层操作，正面或背面 self.ico.DelLayer(self.tmpLays) mes = f'输出目录:{self.filePath},继续将打开PDF' ans = messageBox.showMessage( bitmap='information', title='PDF输出完成', message=mes, buttons=['退出', '继续']) if ans == '继续': os.system(f"/usr/bin/evince {self.filePath} &") self.incam.COM('disp_on') # TODO 转换成png 放到output里面 self.__pdf2PNG() return 0 def __pdf2PNG(self): cmd = f"convert -density 120 -quality 80 -background white -alpha remove {self.filePath} {self.imgPath}" os.system(cmd) def __renderPDF(self, sig_lay): """ 渲染PDF：设置文档结构并构建内容 """ canv = ReportCanvas(self.filePath, pagesize=self.pageSize) self.canvas = canv self.drawingParams = self.__setTemplateParams() # 设置模板的默认参数 self.bgTemp = DrawingTemplate( canv, self.A5[0], self.A5[1], self.drawingParams) lM = 0 rM = 0 tM = 0 bM = 0 self.doc = SimpleDocTemplate(self.filePath, pagesize=self.pageSize, topMargin=tM, bottomMargin=bM, leftMargin=lM, rightMargin=rM, title="MI-13", author=self.userCN)#filePath：最终存放路径； pageSize：画布大小 self.__setPageFrame(self.doc) story = [] story.append(NextPageTemplate('p1')) g2c = self.__createGerber(sig_lay) story.append(FrameBreak()) story.append(g2c) story.append(PageBreak()) self.doc.build(story) # def __setPageFrame(self, doc: SimpleDocTemplate): """设置每一页框架分布 """ frames = [] fh = doc.height / 3 padX = self.drawingParams['padx'] padY = self.drawingParams['pady'] tableFrame = Frame(x1=padX, y1=padY + fh * 2, width=doc.width, height=fh, id='f1') gerberFrame = Frame( x1=padX, y1=padY, width=doc.width, height=fh * 2, id='f2') frames.append(tableFrame) frames.append(gerberFrame) doc.addPageTemplates([PageTemplate(id='p1', frames=frames)]) def getMergeLay(self, lay): """ 将lay备份并将备份层合并为surface """ mergeLay = f'{lay}_merge' self.ico.DelLayer(mergeLay) self.ico.ClearAll() self.ico.DispWork(lay, number=1) self.incam.COM( f'sel_copy_other,dest=layer_name,target_layer={mergeLay},invert=no,dx=0,dy=0,size=0,x_anchor=0,y_anchor=0') self.ico.DispWork(mergeLay, number=1) self.incam.COM( 'sel_cont_resize,accuracy=25.4,break_to_islands=yes,island_size=0,hole_size=0,drill_filter=no,corner_ctl=no') return mergeLay def getDnxSigLayMapping(self, lay: str): """ 获取dnx孔层与其钻带的起始信号层之间的映射关系(1:1) :param dnxLayers:线路层 :return:起始终止是线路层的所有钻孔 """ dnxSigLayMapping = [] for drl in self.ico.GetLayerMatrix()['drlAllLay']: startLay = self.ico.GetLayerMatrix()['drlThrough'][drl]['start'] endLay = self.ico.GetLayerMatrix()['drlThrough'][drl]['end'] if startLay == lay or endLay == lay: dnxSigLayMapping.append(drl) return dnxSigLayMapping # 分析创建光绘的关键部分 def __createGerber(self, sigLay): toRead = [] # 1. 合并信号层为 surface mergeSigLay = self.getMergeLay(sigLay) self.tmpLays.append(mergeSigLay) mergeSigLayFilePath = self.ico.getFeatureFile(self.JOB, self.workStep, mergeSigLay) toRead.append(Feature(mergeSigLayFilePath, layerType='signal')) # 2. 添加钻孔层 drlSet = self.getDnxSigLayMapping(sigLay) for drl in drlSet: drlFilePath = self.ico.getFeatureFile(self.JOB, self.workStep, drl) toRead.append(Feature(drlFilePath, layerType='solder_mask')) # === 创建高亮圆圈层 === highlight_layer = "pofv_highlight_circle" if self.ico.IsLayerExist([highlight_layer]): self.ico.DelLayer([highlight_layer]) self.ico.CreateOrEmptyLay(layer_list=[highlight_layer]) # 获取要高亮的孔坐标（只标第一个） if sigLay not in self.sigDimension or not self.sigDimension[sigLay]['point_x']: # 没有坐标，跳过画圈 pass else: print("111111111111111111111111111") x = float(self.sigDimension[sigLay]['point_x'][0]) y = float(self.sigDimension[sigLay]['point_y'][0]) self.ico.ClearLayer() self.ico.DispWork(highlight_layer) # self.ico.AddPad(x,y,'r120') self.incam.COM(f"add_pad,symbol=r180,polarity=positive,x={x},y={y},mirror=no,angle=0,direction=ccw,resize=0,xscale=1,yscale=1") # self.incam.COM("rv_tab_empty,report=cut_data_rep,is_empty=yes") # self.incam.COM("sel_cut_data,det_tol=25.4,con_tol=25.4,rad_tol=2.54,ignore_width=yes,filter_overlaps=no,delete_doubles=no,use_order=yes,ignore_holes=none,start_positive=yes,polarity_of_touching=same,contourize=yes,simplify=yes,resize_thick_lines=no") # self.incam.COM("rv_tab_view_results_enabled,report=cut_data_rep,is_enabled=no,serial_num=-1,all_count=-1") self.incam.COM("sel_feat2outline,width=3.0,location=on_edge,offset=0.1,polarity=as_feature,keep_original=no,text2limit=no")# 轮廓线 # self.incam.COM(f"display_layer,name={highlight_layer},display=yes,number=3") self.incam.COM("arc2lines,arc_line_tol=1") # self.incam.COM("add_slot,symbol=r20,x_center=0.7836725,y_center=1.85691,len=0.7306875,angle=0.98315767,direction=ccw,dcode=0,drill_type=nplate,attributes=no") # self.incam.COM("add_slot,symbol=r20,x_center=3.7836725,y_center=1.85691,len=0.7306875,angle=0.98315767,direction=ccw,dcode=0,drill_type=nplate,attributes=no") self.ico.DelLayer(highlight_layer + '+++') self.ico.ClearLayer() # sys.exit(0) # 高亮层也导出为 Gerber Feature self.tmpLays.append(highlight_layer) # 确保后续清理" hlight_path = self.ico.getFeatureFile(self.JOB, self.workStep, highlight_layer) toRead.append(Feature(hlight_path, layerType='signal', strokeColor = "#fcfcf6")) # === 计算尺寸和偏移 === unitSizeX, unitSizeY = self.ico.GetStepSize(self.workStep)[0:2] gbWidth = self.doc.width * 0.6 gbHeigth = self.doc.height * 2 / 3 * 0.6 offsetX = self.getOffsetXY(gbWidth, gbHeigth, unitSizeX, unitSizeY, pagesize=( self.doc.width, self.doc.height * 2 / 3))[0] # 创建绘图对象 g2c = Gerber2Canvas( gbWidth, gbHeigth, unitSizeX, unitSizeY, offsetX, 0.1, Origin.leftdown, self.canvas, toRead, rotate=0 ) # === 添加标注箭头=== dimension = [] if sigLay in self.sigDimension and self.sigDimension[sigLay]['point_x']: pointX = self.sigDimension[sigLay]['point_x'][0] pointY = self.sigDimension[sigLay]['point_y'][0] dimensionSingle = Dimension( x0=unitSizeX / 2, y0=unitSizeY * 1.2, x1=float(pointX), y1=float(pointY), direct=Direct.one_arrow, dist='POFV孔铜厚度测量位置', dimColor="#0400FF", # dimLineColor="#0800FF" # ) dimension.append(dimensionSingle) if dimension: g2c.addDimension(dimension) return g2c @staticmethod def getOffsetXY(gbWidth: float, gbHeight: float, unitSizeX: float, unitSizeY: float, pagesize: tuple = (A5[0], A5[1])): """ 获取使Gerber在PDF中居中显示的偏移量 :param gbWidth: pdf中 gerber宽度 :param gbHeight: pdf中 gerber长度 :param unitSizeX: unit宽 :param unitSizeY: unit长 :param pagesize: PDF宽和长 :return: x,y的偏移量 # 创建绘图对象 g2c = Gerber2Canvas(gbWidth,gbHeight,unitSizeX,unitSizeY,offsetX, offsetY, Origin.leftdown,self.canvas,toRead, rotate=0) """ scale1 = math.ceil(gbWidth / unitSizeX) scale2 = math.ceil(gbHeight / unitSizeY) scale = scale1 if scale1 < scale2 else scale2 offsetX = (pagesize[0] - unitSizeX * scale) / (2 * scale) offsetY = (pagesize[1] - unitSizeY * scale) / (2 * scale) return offsetX, offsetY def __setTemplateParams(self): """设置模板的默认参数""" params = {"layer_side": None, "header": "文档密级:内部公开", "footer": "此资料属广芯基板有限公司所有,未经许可,不得扩散.", "note:": "", "tolerance": "", "workstation": "蚀刻开窗-激光钻-电镀-刷板I", "drawing_name": "90874外层最小环宽监控图纸", "jobname": self.jobName, "drawing_no": "MI-13", "drawing_version": self.drawingVer, "units": "mm", "num_page": "", "total_page": self.totalPage, "create": self.userCN, "confirm": "陈伟", "approved": "刘丹洪", "padx": 5 * mm, "pady": 5 * mm, } return params def drawBackground(self, canv: ReportCanvas, doc: SimpleDocTemplate): """ 画页眉页脚用的函数 """ num = canv.getPageNumber() if num == 1: face = 'Unit Top side' else: face = 'Unit Bottom side' params = self.drawingParams params["num_page"] = str(num) params['layer_side'] = face bgTemp = DrawingTemplate(canv, A5[0], A5[1], params) bgTemp.parser() if __name__ == "__main__": app = QApplication(sys.argv) analyzer = Dr_POFV_Map() 上述代码在本地电脑可以正常运行，但是在另一个电脑报错： gen_line-8007-Field does not exist The command sel_copy_other does not have the field subsytem