Comments (5)
@Paleve can you provide more detail about the error
from text-detection-ctpn.
@eragonruan 前面权重载入是没问题的,当前向计算的时候会报总线错误
2017-11-15 09:50:54.910148: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.
2017-11-15 09:50:54.910152: W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.
Tensor("Placeholder:0", shape=(?, ?, ?, 3), dtype=float32)
Tensor("conv5_3/conv5_3:0", shape=(?, ?, ?, 512), dtype=float32)
Tensor("rpn_conv/3x3/rpn_conv/3x3:0", shape=(?, ?, ?, 512), dtype=float32)
WARNING:tensorflow:<tensorflow.python.ops.rnn_cell_impl.BasicLSTMCell object at 0x10f73c290>: Using a concatenated state is slower and will soon be deprecated. Use state_is_tuple=True.
Tensor("lstm_o/Reshape:0", shape=(?, ?, ?, 128), dtype=float32)
Tensor("lstm_o/Reshape:0", shape=(?, ?, ?, 128), dtype=float32)
Tensor("rpn_cls_score/Reshape:0", shape=(?, ?, ?, 20), dtype=float32)
Tensor("rpn_cls_prob:0", shape=(?, ?, ?, ?), dtype=float32)
Tensor("Reshape_5:0", shape=(?, ?, ?, 20), dtype=float32)
Tensor("rpn_bbox_pred/Reshape:0", shape=(?, ?, ?, 40), dtype=float32)
Tensor("Placeholder_1:0", shape=(?, 3), dtype=float32)
Loading network VGGnet_test... Restoring from checkpoints/VGGnet_fast_rcnn_iter_50000.ckpt... done
done.
Bus error: 10
from text-detection-ctpn.
@Paleve This may caused by your system, not the code. you can find more in these two issue: tensorflow/issues/7663, tensorflow/issues/6744
from text-detection-ctpn.
@eragonruan thanks a lot for your reply, however I meet a new issue , the loss is strange in the training. The original lose is around 2.0, but it increases to 50+ sometimes in the training, I think it's because the box isn't uniform splitted into 16 pixels' proposals, did you meet the similar problem before? I use ICDAR2017 MLT dataset. What's more, the box loss decrease slowly.
from text-detection-ctpn.
@Paleve no, but sometimes the loss may jump to 8~10, this is caused by some "bad" images, because texts in mlt dataset contains oriented texts, if the rotated angle is too large, the loss may jump. box loss is hard to converge. you need to adjust the learning rate when cls loss has converged, and eventually box loss can reach to about 0.1 after 50k iterations in my experiment.
from text-detection-ctpn.
Related Issues (20)
- 百度网盘的权重链接失效了,有哪位大佬能提供下训练好的权重文件吗? HOT 4
- training with images from google
- demo.py中 将box转换为原始尺寸时,代码可能有一点小错误
- Crop detected region with result from .txt file but I got the wrong region? HOT 1
- loss值不下降,检测为0
- 请问有data 数据的百度云链接吗? 过期了 HOT 3
- It does not work with tensorflow 2.3.1
- bug in utils.py
- Resource Exhausted: OOM issue
- how to create our own training model? HOT 1
- when i train model My gpu utilization is nearly 0
- ImportError: cannot import name 'bbox' from partially initialized module 'lib.utils' (most likely due to a circular import) (/home/barry/Github/chinese_ocr/ctpn/lib/utils/__init__.py) HOT 3
- 如何预测anchor在X方向上的偏移?有人实现了么,当前模型检测出的文本行,两边会有多余的边缘,在识别的时候,多出的部分严重影响识别效果,
- 运行spilt_label.py结果一直显示Error processing HOT 1
- File "nms.pyx", line 28, in bbox.nms.nms ValueError: Buffer dtype mismatch, expected 'intp_t' but got 'long'
- there are some problem in check[oints HOT 1
- Model Quantization
- 请问怎样把检测出来的长文本进行字符切割?
- Required Python Version 2.7
- during multi-language model training, accuracy of English/numbers is worse than Chinese characters, how to improve?
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from text-detection-ctpn.