QA Official

Keras Beginner-Level (2)-Sparrows, though small, have all five organs

https://qaofficial.com/post/2019/03/31/23944-keras-beginner-level-2-sparrows-though-small-have-all-five-organs.html 2019-03-31
1. What is Keras I don't know when, suddenly confused about what Keras is.Keras is also said to be a brass instrument. In fact, Keras's name comes from the Horn Gate in the Greek classical epic Odyssey. It is the place where real things enter and exit dreams and reality.The Odyssey says that the Ivory Gate is only a dream that cannot be fulfilled. Only those who enter the Horn

Label Classification

https://qaofficial.com/post/2019/03/31/31490-label-classification.html 2019-03-31
tag elements are generally divided into three types: block elements, inline elements (inline elements) and inline block elements.1, massive elements are: 、、...、、、、、、 、 2, inline elements are: 、、、、、、、、、、3, inline block elements are: 、

Overview: RNN Application in Target Recognition in Computer Vision

https://qaofficial.com/post/2019/03/31/23990-overview-rnn-application-in-target-recognition-in-computer-vision.html 2019-03-31
Deep learning has made great progress in the field of computer vision. In recent years CNN has been the framework adopted by current mainstream models.In the past six months, RNN/LSTM application has gradually become a trend in the field of identification. RNN has unique advantages over CNN in the context of obtaining targets.Following is an analysis of recent articles on RNN for target recognition. 1、Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks CVPR2016

Target Detection-Faster R-CNN Training Process Source Code Understanding

https://qaofficial.com/post/2019/03/31/23889-target-detection-faster-r-cnn-training-process-source-code-understanding.html 2019-03-31
Faster R-CNN Training Process Source Code Understanding training script./tools/train_net.py main function starts. Data Reading Layer RoIDataLayer First of all, imdb, roidb = combined_roidb(args.imdb_name) # 输入参数 imdb_name,默认是 voc_2007_trainval(数据集名字) print '{:d} roidb entries'.format(len(roidb)) Then,

U-Net Network Interpretation for Medical Image Segmentation

https://qaofficial.com/post/2019/03/31/23819-u-net-network-interpretation-for-medical-image-segmentation.html 2019-03-31
U-Net original: u-net: convolutional networks for biomedical image segmentation TensorFlow implementation: jakret/tf _ unet  

Use of Pre-trained Word Vector in Keras Model

https://qaofficial.com/post/2019/03/31/24025-use-of-pre-trained-word-vector-in-keras-model.html 2019-03-31
From: wuwt.me/2017/08/21/pre-trained-embedding-keras/ code download: https://github.com/keras-team/keras/blob/2.0.5/examples/pretrain _ word _ embeddings.py " word vector" (word embedding) is a natural language processing technology that maps the semantics of a class of words into vector space.That is, a word is represented by a specific vector. The distance between vectors (for example, L2 normal form distance or cosine distance more commonly used between any two vectors) represents the semantic relationship between words to some extent.The

Using Tensorflow to Realize Picture Classification+Detailed Annotation

https://qaofficial.com/post/2019/03/31/23794-using-tensorflow-to-realize-picture-classification-detailed-annotation.html 2019-03-31
Use tensorflow to classify flowers. The training samples are online flower data sets, and the test samples are pictures of various flowers downloaded again from the Internet. The code is as follows. from skimage import io,transform #skimage模块下的io transform(图像的形变与缩放)模

ai from codecs-sentiment classification (LSTM on TFlearn)

https://qaofficial.com/post/2019/03/31/24030-ai-from-codecs-sentiment-classification-lstm-on-tflearn.html 2019-03-31
Preface this article will explain how to use LSTM to classify text emotion based on TFlearn from a code perspective.If you are not familiar with TFlearn and LSTM, it doesn't matter, first look at the code (use LSTM to classify the IMDB dataset emotionally).From the point of view of code is very simple, so even if not familiar with, look at the code, when the code is already familiar with

how to use data generator in Keras

https://qaofficial.com/post/2019/03/31/23695-how-to-use-data-generator-in-keras.html 2019-03-31
motivation If you need to train a model that uses large data sets, you will probably encounter a problem: too large data sets cannot be loaded into memory at the same time, but the performance of the model has to use large data sets, so how to call large data sets in the training process is a very realistic problem.In this blog post, we will introduce a batch-by-batch data set generation method to generate real-time multi-core data sets and deliver the in-depth learning model to you immediately.

implement simple sequence2sequence model based on LSTM

https://qaofficial.com/post/2019/03/31/24012-implement-simple-sequence2sequence-model-based-on-lstm.html 2019-03-31
(1) Prepare data batch_size = 1000 # Batch size for training. epochs = 50 # Number of epochs to train for. latent_dim = 256 # Latent dimensionality of the encoding space. num_samples = 10000 # Number of samples to train on. # Path to the data txt file on disk. data_path = '../../input_data/fra-eng/fra.txt' # Vectorize the data. input_texts = [] target_texts = [] input_characters = set() target_characters = set() with open(data_path, 'r', encoding='utf-8') as f: lines = f.