mother and baby

import java.io.FileInputStream; 2 import java.io.FileOutputStream; 3 import java.io.IOException; 4 import java.io.InputStream; 5 import java.io.OutputStream; 6 7 import org.apache.commons.codec.binary.Base64; 8 9 10 /** 11 * 将图片转换为Base64<br> 12 * 将base64编码字符串解码成img图片 13 * @创建时间 2015-06-01 15:50 14 * 15 */ 16 public class Img2Base64Util { 17 18 public static void main(String[] args) { 19 String imgFile = "d:\\3.jpg";//待处理的图片 20 String imgbese=getImgStr(imgFile); 21 System.out.println(imgbese.length()); 22 System.out.println(imgbese); 23 String imgFilePath = "d:\\332.jpg";//新生成的图片 24 generateImage(imgbese,imgFilePath); 25 } 26 /** 27 * 将图片转换成Base64编码 28 * @param imgFile 待处理图片 29 * @return 30 */ 31 public static String getImgStr(String imgFile){ 32 //将图片文件转化为字节数组字符串，并对其进行Base64编码处理 33 34 35 InputStream in = null; 36 byte[] data = null; 37 //读取图片字节数组 38 try 39 { 40 in = new FileInputStream(imgFile); 41 data = new byte[in.available()]; 42 in.read(data); 43 in.close(); 44 } 45 catch (IOException e) 46 { 47 e.printStackTrace(); 48 } 49 return new String(Base64.encodeBase64(data)); 50 } 51 52 /** 53 * 对字节数组字符串进行Base64解码并生成图片 54 * @param imgStr 图片数据 55 * @param imgFilePath 保存图片全路径地址 56 * @return 57 */ 58 public static boolean generateImage(String imgStr,String imgFilePath){ 59 // 60 if (imgStr == null) //图像数据为空 61 return false; 62 63 try 64 { 65 //Base64解码 66 byte[] b = Base64.decodeBase64(imgStr); 67 for(int i=0;i<b.length;++i) 68 { 69 if(b[i]<0) 70 {//调整异常数据 71 b[i]+=256; 72 } 73 } 74 //生成jpeg图片 75 76 OutputStream out = new FileOutputStream(imgFilePath); 77 out.write(b); 78 out.flush(); 79 out.close(); 80 return true; 81 } 82 catch (Exception e) 83 { 84 return false; 85 } 86 } 87 }

prompt angel of compassion

有个问题很疑惑，烧录文件方式制卡的话，图片1出的红框的SD卡应该是张空卡；那图片中前提条件的第三个点是在没有SD卡系统的情况下登录吗

1. 系统环境硬件环境(Ascend/GPU/CPU): modelart软件环境:– MindSpore 版本: 1.5.1执行模式：动态图(PYNATIVE_MODE) – Python 版本: 3.7.6– 操作系统平台: linux2. 问题描述2.1 问题描述YOLOv3_darknet53图片解码失败2.2 报错信息2.3 脚本代码# Copyright 2020-2022 Huawei Technologies Co., Ltd # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================ """YoloV3 train.""" import os import time import datetime import mindspore as ms import mindspore.nn as nn import mindspore.communication as comm from src.yolo import YOLOV3DarkNet53, YoloWithLossCell from src.logger import get_logger from src.util import AverageMeter, get_param_groups, cpu_affinity from src.lr_scheduler import get_lr from src.yolo_dataset import create_yolo_dataset from src.initializer import default_recurisive_init, load_yolov3_params from src.util import keep_loss_fp32 from model_utils.config import config # only useful for huawei cloud modelarts. from model_utils.moxing_adapter import moxing_wrapper, modelarts_pre_process ms.set_seed(1) def conver_training_shape(args): training_shape = [int(args.training_shape), int(args.training_shape)] return training_shape def set_graph_kernel_context(): if ms.get_context("device_target") == "GPU": ms.set_context(enable_graph_kernel=True) ms.set_context(graph_kernel_flags="--enable_parallel_fusion " "--enable_trans_op_optimize " "--disable_cluster_ops=ReduceMax,Reshape " "--enable_expand_ops=Conv2D") def network_init(args): device_id = int(os.getenv('DEVICE_ID', '0')) ms.set_context(mode=ms.GRAPH_MODE, device_target=args.device_target, save_graphs=False, device_id=device_id) set_graph_kernel_context() # Set mempool block size for improving memory utilization, which will not take effect in GRAPH_MODE if ms.get_context("mode") == ms.PYNATIVE_MODE: ms.set_context(mempool_block_size="31GB") # Since the default max memory pool available size on ascend is 30GB, # which does not meet the requirements and needs to be adjusted larger. if ms.get_context("device_target") == "Ascend": ms.set_context(max_device_memory="31GB") profiler = None if args.need_profiler: profiling_dir = os.path.join("profiling", datetime.datetime.now().strftime('%Y-%m-%d_time_%H_%M_%S')) profiler = ms.profiler.Profiler(output_path=profiling_dir) # init distributed if args.is_distributed: comm.init() args.rank = comm.get_rank() args.group_size = comm.get_group_size() if args.device_target == "GPU" and args.bind_cpu: cpu_affinity(args.rank, min(args.group_size, args.device_num)) # select for master rank save ckpt or all rank save, compatible for model parallel args.rank_save_ckpt_flag = 0 if args.is_save_on_master: if args.rank == 0: args.rank_save_ckpt_flag = 1 else: args.rank_save_ckpt_flag = 1 # logger args.outputs_dir = os.path.join(args.ckpt_path, datetime.datetime.now().strftime('%Y-%m-%d_time_%H_%M_%S')) args.logger = get_logger(args.outputs_dir, args.rank) args.logger.save_args(args) return profiler def parallel_init(args): ms.reset_auto_parallel_context() parallel_mode = ms.ParallelMode.STAND_ALONE degree = 1 if args.is_distributed: parallel_mode = ms.ParallelMode.DATA_PARALLEL degree = comm.get_group_size() ms.set_auto_parallel_context(parallel_mode=parallel_mode, gradients_mean=True, device_num=degree) @moxing_wrapper(pre_process=modelarts_pre_process) def run_train(): """Train function.""" if config.lr_scheduler == 'cosine_annealing' and config.max_epoch > config.T_max: config.T_max = config.max_epoch config.lr_epochs = list(map(int, config.lr_epochs.split(','))) config.data_root = os.path.join(config.data_dir, 'train2014') config.annFile = os.path.join(config.data_dir, 'annotations/instances_train2014.json') profiler = network_init(config) loss_meter = AverageMeter('loss') parallel_init(config) network = YOLOV3DarkNet53(is_training=True) # default is kaiming-normal default_recurisive_init(network) load_yolov3_params(config, network) network = YoloWithLossCell(network) config.logger.info('finish get network') if config.training_shape: config.multi_scale = [conver_training_shape(config)] ds = create_yolo_dataset(image_dir=config.data_root, anno_path=config.annFile, is_training=True, batch_size=config.per_batch_size, device_num=config.group_size, rank=config.rank, config=config) config.logger.info('Finish loading dataset') config.steps_per_epoch = ds.get_dataset_size() lr = get_lr(config) opt = nn.Momentum(params=get_param_groups(network), momentum=config.momentum, learning_rate=ms.Tensor(lr), weight_decay=config.weight_decay, loss_scale=config.loss_scale) is_gpu = ms.get_context("device_target") == "GPU" if is_gpu: loss_scale_value = 1.0 loss_scale = ms.FixedLossScaleManager(loss_scale_value, drop_overflow_update=False) network = ms.build_train_network(network, optimizer=opt, loss_scale_manager=loss_scale, level="O2", keep_batchnorm_fp32=False) keep_loss_fp32(network) else: network = nn.TrainOneStepCell(network, opt, sens=config.loss_scale) network.set_train() t_end = time.time() data_loader = ds.create_dict_iterator(output_numpy=True) first_step = True stop_profiler = False for epoch_idx in range(config.max_epoch): for step_idx, data in enumerate(data_loader): images = data["image"] input_shape = images.shape[2:4] config.logger.info('iter[{}], shape{}'.format(step_idx, input_shape[0])) images = ms.Tensor.from_numpy(images) batch_y_true_0 = ms.Tensor.from_numpy(data['bbox1']) batch_y_true_1 = ms.Tensor.from_numpy(data['bbox2']) batch_y_true_2 = ms.Tensor.from_numpy(data['bbox3']) batch_gt_box0 = ms.Tensor.from_numpy(data['gt_box1']) batch_gt_box1 = ms.Tensor.from_numpy(data['gt_box2']) batch_gt_box2 = ms.Tensor.from_numpy(data['gt_box3']) loss = network(images, batch_y_true_0, batch_y_true_1, batch_y_true_2, batch_gt_box0, batch_gt_box1, batch_gt_box2) loss_meter.update(loss.asnumpy()) # it is used for loss, performance output per config.log_interval steps. if (epoch_idx * config.steps_per_epoch + step_idx) % config.log_interval == 0: time_used = time.time() - t_end if first_step: fps = config.per_batch_size * config.group_size / time_used per_step_time = time_used * 1000 first_step = False else: fps = config.per_batch_size * config.log_interval * config.group_size / time_used per_step_time = time_used / config.log_interval * 1000 config.logger.info('epoch[{}], iter[{}], {}, fps:{:.2f} imgs/sec, ' 'lr:{}, per step time: {}ms'.format(epoch_idx + 1, step_idx + 1, loss_meter, fps, lr[step_idx], per_step_time)) t_end = time.time() loss_meter.reset() if config.need_profiler: if epoch_idx * config.steps_per_epoch + step_idx == 10: profiler.analyse() stop_profiler = True break if config.rank_save_ckpt_flag: ckpt_path = os.path.join(config.outputs_dir, 'ckpt_' + str(config.rank)) if not os.path.exists(ckpt_path): os.makedirs(ckpt_path, exist_ok=True) ckpt_name = os.path.join(ckpt_path, "yolov3_{}_{}.ckpt".format(epoch_idx + 1, config.steps_per_epoch)) ms.save_checkpoint(network, ckpt_name) if stop_profiler: break config.logger.info('==========end training===============') if __name__ == "__main__": run_train()3. 根因分析记录一下排查流程吧，用户使用了自有数据集，因此无法判断是代码异常还是数据集异常，先在相同的网络上跑coco公共数据集，发现不报错，排除代码问题，然后使用get_batch_size()，get_class_indexing()，get_col_names()，get_dataset_size()，get_repeat_count()，查看数据集是否争取加载，发现正确加载。排除图片问题。一次偶然的机会，偶然发现其实是数据集中描述图片标签的json文件损坏，某张图片未查找到，至此，问题排查结束。4. 解决方案重新换了一份json文件正确的数据集（非公开的）。5. 经验总结排查数据集问题，不止要注意对数据集目录结构、图像的问题排查，保存图像信息的json也很重要。

请问一下，图片审核异步提交的任务华为云这边保留多久，我这边很多超过半天的或者一天的任务去查就是404没找到任务

【Atlas200产品】【PCB设计】 原理图，申请帮忙审查，邮件已发送，谢谢请重点审查以下几个网络及其外围电路和第5页的电平转换电路。1. GE_PHY_RST_N2. GE_PHY_INT3. CLKREQB4. LANWAKEB5. HOST_RST_N6. PCIE_PERST_N7. ETH0_INTB8. GPIO_739. PERSTB10. PHYRSTB

您好，HOST_RST_N这个管脚的复位电路有没有参考设计可以参考呀

您好，我用的PHY芯片是RTL8211F,其中，中断31脚手册 上写的是上拉3.3V，但是在Atlas200 的中断输入脚是1.8V的，可以把PHY芯片的31脚上拉3.3后接到Atlas的GE_PHY_INT输入脚吗​

【Atlas200产品】【PCB设计】 原理图，申请帮忙审核，邮件已发送，谢谢

之前发的审核帖子找不到了怎么办呀，重新发一个邮件和帖子吗

内容审核-图像有以下应用场景：视频直播在互动直播场景中，成千上万个房间并发直播，人工审核直播内容几乎不可能。基于图像审核能力，可对所有房间内容实时监控，识别可疑房间并进行预警。场景优势如下：准确率高：基于改进的深度学习算法，检测准确率高。响应速度快：视频直播响应速度速度小于0.1秒。在线商城智能审核商家/用户上传图像，高效识别并预警不合规图片，防止涉黄、涉暴、政治敏感类图像发布，降低人工审核成本和业务违规风险。场景优势如下：准确率高：基于改进的深度学习算法，检测准确率高。响应速度快：单张图像识别速度小于0.1秒。网站论坛不合规图片的识别和处理是用户原创内容（UGC）类网站的重点工作，基于内容审核，可以识别并预警用户上传的不合规图片，帮助客户快速定位处理，降低业务违规风险。场景优势如下：准确率高：基于改进的深度学习算法，检测准确率高。响应速度快：单张图像识别速度小于0.1秒。