HandsLing

Results 24 comments of HandsLing

Thank U for creating Docface！got a good results after finetuned!Can u tell how to save model as SavedModel?tf的那个savedModel格式，搞了好久不成功

@eguoguo321 请问您是用什么数据去finetune的呢？自己制作的数据集吗？

where is the cython_bbox file?

i did that but not have cython_bbox,just bbox,so i changed cython_bbox into bbox in where it was imported

evaluation

i follow your steps,but the output has three members: easy,hard.. they are all zeros,just has the time

evaluation

2017-05-22 19:11:08,795 root INFO Raw Results: 2017-05-22 19:11:08,795 root INFO val easy (raw) : 0.0000 2017-05-22 19:11:08,795 root INFO val medium (raw) : 0.0000 2017-05-22 19:11:08,795 root INFO val hard...

关于运行时间和卷积层

你指的是把kitti提供的训练集拆分成两半的那种吗？我是直接把原来的训练集复制一份，最后根据train.txt或者是val.txt中的数据来删除数据得到训练集和测试集

关于运行时间和卷积层

import os import glob import numpy as np path = '/home/lingck/Kitti/object/train.txt' train_dir = '/home/lingck/second.pytorch/second/object/testing' f = open(path,'r').readlines() r = [] for ff in f: r.append(ff.split('\n')[0].zfill(6)) f_rgb = glob.glob(os.path.join(train_dir, 'image_2', '*.png'))...

训练的时候train_accuracy一直下降

@wubaoyuan 求解答。谢谢，另外我的tfrecord是自己生成的，我是二分类，但是最后只有一个tfrecord文件

两张照片不同的人，距离也很小

@qwn19970728 thx, but i did changed that before, and i can get the same embedding for one image twice. and the distance in this issue was calculated by these embeddings

有几个问题想请教一下

然后我把剪枝后的模型和剪枝前的模型对比了一下，速度上剪枝后的模型还慢些了，我看了模型结构好像是比官方的yolo模型多了很多的bn结

请问支持流式语音识别再训练吗？

那训练完之后，怎么把训练完的模型替换到流式语音识别部署那里去呢？替换一个model.yaml已经对应的模型就可以了么