- hw0:文献阅读(10篇)
- hw1:MNIST数据集手写数字识别(准确率98.83%)
- hw2:猫狗二分类(准确率91%以上)
- hw3:自动写诗(首句写诗,藏头写诗)
- hw4:电影中文影评情感分类(准确率,精度,召回率,F1分数,混淆矩阵指标分别达到了:86.18%,0.8626,0.8579,0.8603,[[157.0, 26.0], [25.0, 161.0]])
- hw5:车牌识别,省份、地区、5个字符,3部分验证集准确率分别达到 100%,100%,98.5%。
- hw6:实现 Transformer 机器翻译,英译汉。 https://github.com/Allenem/transformer
由于猫狗分类、维基百科中文word2vec.bin等数据文件较大,本人并未全部上传至此仓库,可在如下地址下载数据集。
- 实验1数据:MINIST 手写数据集
- 实验2数据:从kaggle比赛官网 下载所需的猫狗分类数据;或者直接从此下载训练集和测试集
- 实验3数据:
tang.npz
文件较小,已上传至本仓库,或者:链接:https://pan.baidu.com/s/1dtf9HOEY1jzqR51tyyQfjg 提取码:65qv 复制这段内容后打开百度网盘手机App,操作更方便哦 - 实验4数据:
Dataset
(包含test.txt
,train.txt
,validation.txt
,wiki_word2vec_50.bin
):链接:https://pan.baidu.com/s/1VDYXwjSLO1sTC0XJKq9ggA 提取码:buuq 复制这段内容后打开百度网盘手机App,操作更方便哦 - 实验5数据:
LPD_dataset
原始街道拍的含车牌汽车图像,dataset-train&val
韩train和validation两个文件夹,分别包含province、area、letter三个文件夹,包含将车牌分割好的单个字块。PROVINCES = ("沪", "京", "闽", "苏", "粤", "浙"), AREAS = ("A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z"), LETTERS = ("0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "A", "B", "C", "D", "E", "F", "G", "H", "J", "K", "L", "M", "N", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z")链接:https://pan.baidu.com/s/192uU_MrW2kOCVzfocR6zWg 提取码:3ry8
以下结构为完整结构:
.
│ README.md
│
├─hw0_paper_reading
│ 0Deep residual learning for image recognition.pdf
│ 0图像识别的深度残差学习.pptx
│ 1Long-term Recurrent Convolutional Networks for Visual Recognition and Description.pdf
│ 1用于视觉识别和描述的长期循环卷积网络.pptx
│ 2Speech recognition with deep recurrent neural networks.pdf
│ 2基于深度循环神经网络的语音识别.pptx
│ 3Attention Is All You Need.pdf
│ 3你只需要注意力.pptx
│ 4A Reinforced Topic-Aware Convolutional Sequence-to-Sequence Model for Abstractive Text Summarization.pdf
│ 4一种增强主题感知的卷积序列对序列的摘要模型.pptx
│ 5Adaptive RNN for target-dependent Twitter Sentiment Classification.pdf
│ 5自适应递归神经网络用于目标相关的推特情感分类.pptx
│ 6Convolutional Neural Networks for Sentence Classification.pdf
│ 6用于句子分类的卷积神经网络.pptx
│ 7Fast R-CNN.pdf
│ 7快速区域卷积神经网络.pptx
│ 8You Only Look Once-Unified, Real-Time Object Detection.pdf
│ 8你只需看一次——统一的实时的目标检测.pptx
│ 9YOLOv3- An Incremental Improvement.pdf
│ 9YOLOv3——一个渐进的改进.pptx
│
├─hw1_writing_digit_recognition
│ │ index.py
│ │ OUTPUT.txt
│ │ 实验报告.docx
│ │ 实验报告.pptx
│ │
│ └─MNIST_data
│ t10k-images-idx3-ubyte.gz
│ t10k-labels-idx1-ubyte.gz
│ train-images-idx3-ubyte.gz
│ train-labels-idx1-ubyte.gz
│
├─hw2_catVSdog_classification
│ │ index.py
│ │ OUTPUT.txt
│ │ 实验报告.docx
│ │ 实验报告.pptx
│ │
│ └─data
│ ├─test
│ │ ├─cat
│ │ │ cat.12000.jpg
│ │ │ ...
│ │ │ cat.12499.jpg
│ │ │
│ │ └─dog
│ │ dog.12000.jpg
│ │ ...
│ │ dog.12499.jpg
│ │
│ └─train
│ ├─cat
│ │ cat.0.jpg
│ │ ...
│ │ cat.999.jpg
│ │
│ └─dog
│ dog.0.jpg
│ ...
│ dog.999.jpg
│
├─hw3_write_poetry_automatically
│ aotomatic_writing_poetry.ipynb
│ model.pth
│ tang.npz
│ 实验报告.docx
│ 实验报告.pptx
│
├─hw4_movie_chinese_comments_sentiment_classification
│ │ Chinese_movie_comments_sentiment_classification.ipynb
│ │ model.pth
│ │ 实验报告.docx
│ │ 实验报告.pptx
│ │
│ └─Dataset
│ test.txt
│ train.txt
│ validation.txt
│ wiki_word2vec_50.bin
│
└─hw5_vehicle_license_plate_recognition
│ preprocessing.py
│ train-license-province.py
│ train-license-area.py
│ train-license-letter.py
│ 实验报告.docx
│ 实验报告.pptx
│
├─LPD_dataset
│ ├─train
│ └─val
│
├─preprocessed
│ ├─train
│ │ ├─correct
│ │ ├─crop
│ │ │ ├─川A09X20
...
│ │ │ └─粤BA103N
│ │ └─rgb2gray
│ └─val
│ ├─correct
│ ├─crop
│ │ ├─浙A03168
...
│ │ └─粤X30479
│ └─rgb2gray
│
├─dataset-train&val
│ ├─training-set
│ │ ├─area
│ │ │ ├─10
...
│ │ │ └─35
│ │ ├─letter
│ │ │ ├─0
...
│ │ │ ├─33
│ │ └─province
│ │ ├─0
...
│ │ └─5
│ └─validation-set
│ ├─area
│ │ ├─10
...
│ │ └─35
│ ├─letter
│ │ ├─0
...
│ │ ├─33
│ └─province
│ ├─0
...
│ └─5
│
├─test_images
└─train-saver
├─area
├─letter
└─province